diff --git "a/wandb/run-20220616_112025-1526e2dp/files/config.yaml" "b/wandb/run-20220616_112025-1526e2dp/files/config.yaml" new file mode 100644--- /dev/null +++ "b/wandb/run-20220616_112025-1526e2dp/files/config.yaml" @@ -0,0 +1,10058 @@ +wandb_version: 1 + +_n_gpu: + desc: null + value: 1 +_name_or_path: + desc: null + value: facebook/wav2vec2-xls-r-1b +_wandb: + desc: null + value: + cli_version: 0.12.15 + framework: huggingface + huggingface_version: 4.18.0 + is_jupyter_run: false + is_kaggle_kernel: false + m: + - 1: train/global_step + 6: + - 3 + - 1: gradients/lm_head\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/lm_head\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/lm_head\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/lm_head\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/lm_head\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/lm_head\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.pos_conv_embed\.conv\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.pos_conv_embed\.conv\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.pos_conv_embed\.conv\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.pos_conv_embed\.conv\.weight_v._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.pos_conv_embed\.conv\.weight_v.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.pos_conv_embed\.conv\.weight_v.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.pos_conv_embed\.conv\.weight_g._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.pos_conv_embed\.conv\.weight_g.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.pos_conv_embed\.conv\.weight_g.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.masked_spec_embed._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.masked_spec_embed.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.masked_spec_embed.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.feature_projection\.projection\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.feature_projection\.projection\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.feature_projection\.projection\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.feature_projection\.projection\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.feature_projection\.projection\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.feature_projection\.projection\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.feature_projection\.layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.feature_projection\.layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.feature_projection\.layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.feature_projection\.layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.feature_projection\.layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.feature_projection\.layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.11\.feed_forward\.output_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.11\.feed_forward\.output_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.11\.feed_forward\.output_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.11\.feed_forward\.output_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.11\.feed_forward\.output_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.11\.feed_forward\.output_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.11\.feed_forward\.intermediate_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.11\.feed_forward\.intermediate_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.11\.feed_forward\.intermediate_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.11\.feed_forward\.intermediate_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.11\.feed_forward\.intermediate_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.11\.feed_forward\.intermediate_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.11\.final_layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.11\.final_layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.11\.final_layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.11\.final_layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.11\.final_layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.11\.final_layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.11\.attention\.out_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.11\.attention\.out_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.11\.attention\.out_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.11\.attention\.out_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.11\.attention\.out_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.11\.attention\.out_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.11\.attention\.v_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.11\.attention\.v_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.11\.attention\.v_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.11\.attention\.v_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.11\.attention\.v_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.11\.attention\.v_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.11\.attention\.k_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.11\.attention\.k_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.11\.attention\.k_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.11\.attention\.k_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.11\.attention\.k_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.11\.attention\.k_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.11\.attention\.q_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.11\.attention\.q_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.11\.attention\.q_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.11\.attention\.q_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.11\.attention\.q_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.11\.attention\.q_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.11\.layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.11\.layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.11\.layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.11\.layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.11\.layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.11\.layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.3\.feed_forward\.output_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.3\.feed_forward\.output_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.3\.feed_forward\.output_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.3\.feed_forward\.output_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.3\.feed_forward\.output_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.3\.feed_forward\.output_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.3\.feed_forward\.intermediate_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.3\.feed_forward\.intermediate_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.3\.feed_forward\.intermediate_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.3\.feed_forward\.intermediate_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.3\.feed_forward\.intermediate_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.3\.feed_forward\.intermediate_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.3\.final_layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.3\.final_layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.3\.final_layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.3\.final_layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.3\.final_layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.3\.final_layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.3\.attention\.out_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.3\.attention\.out_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.3\.attention\.out_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.3\.attention\.out_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.3\.attention\.out_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.3\.attention\.out_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.3\.attention\.v_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.3\.attention\.v_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.3\.attention\.v_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.3\.attention\.v_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.3\.attention\.v_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.3\.attention\.v_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.3\.attention\.k_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.3\.attention\.k_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.3\.attention\.k_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.3\.attention\.k_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.3\.attention\.k_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.3\.attention\.k_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.3\.attention\.q_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.3\.attention\.q_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.3\.attention\.q_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.3\.attention\.q_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.3\.attention\.q_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.3\.attention\.q_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.3\.layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.3\.layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.3\.layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.3\.layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.3\.layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.3\.layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.46\.feed_forward\.output_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.46\.feed_forward\.output_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.46\.feed_forward\.output_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.46\.feed_forward\.output_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.46\.feed_forward\.output_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.46\.feed_forward\.output_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.46\.feed_forward\.intermediate_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.46\.feed_forward\.intermediate_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.46\.feed_forward\.intermediate_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.46\.feed_forward\.intermediate_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.46\.feed_forward\.intermediate_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.46\.feed_forward\.intermediate_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.46\.final_layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.46\.final_layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.46\.final_layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.46\.final_layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.46\.final_layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.46\.final_layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.46\.attention\.out_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.46\.attention\.out_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.46\.attention\.out_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.46\.attention\.out_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.46\.attention\.out_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.46\.attention\.out_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.46\.attention\.v_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.46\.attention\.v_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.46\.attention\.v_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.46\.attention\.v_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.46\.attention\.v_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.46\.attention\.v_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.46\.attention\.k_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.46\.attention\.k_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.46\.attention\.k_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.46\.attention\.k_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.46\.attention\.k_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.46\.attention\.k_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.46\.attention\.q_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.46\.attention\.q_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.46\.attention\.q_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.46\.attention\.q_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.46\.attention\.q_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.46\.attention\.q_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.46\.layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.46\.layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.46\.layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.46\.layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.46\.layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.46\.layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.42\.feed_forward\.output_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.42\.feed_forward\.output_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.42\.feed_forward\.output_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.42\.feed_forward\.output_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.42\.feed_forward\.output_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.42\.feed_forward\.output_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.42\.feed_forward\.intermediate_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.42\.feed_forward\.intermediate_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.42\.feed_forward\.intermediate_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.42\.feed_forward\.intermediate_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.42\.feed_forward\.intermediate_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.42\.feed_forward\.intermediate_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.42\.final_layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.42\.final_layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.42\.final_layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.42\.final_layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.42\.final_layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.42\.final_layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.42\.attention\.out_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.42\.attention\.out_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.42\.attention\.out_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.42\.attention\.out_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.42\.attention\.out_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.42\.attention\.out_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.42\.attention\.v_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.42\.attention\.v_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.42\.attention\.v_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.42\.attention\.v_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.42\.attention\.v_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.42\.attention\.v_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.42\.attention\.k_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.42\.attention\.k_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.42\.attention\.k_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.42\.attention\.k_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.42\.attention\.k_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.42\.attention\.k_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.42\.attention\.q_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.42\.attention\.q_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.42\.attention\.q_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.42\.attention\.q_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.42\.attention\.q_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.42\.attention\.q_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.42\.layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.42\.layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.42\.layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.42\.layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.42\.layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.42\.layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.15\.feed_forward\.output_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.15\.feed_forward\.output_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.15\.feed_forward\.output_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.15\.feed_forward\.output_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.15\.feed_forward\.output_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.15\.feed_forward\.output_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.15\.feed_forward\.intermediate_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.15\.feed_forward\.intermediate_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.15\.feed_forward\.intermediate_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.15\.feed_forward\.intermediate_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.15\.feed_forward\.intermediate_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.15\.feed_forward\.intermediate_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.15\.final_layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.15\.final_layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.15\.final_layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.15\.final_layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.15\.final_layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.15\.final_layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.15\.attention\.out_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.15\.attention\.out_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.15\.attention\.out_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.15\.attention\.out_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.15\.attention\.out_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.15\.attention\.out_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.15\.attention\.v_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.15\.attention\.v_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.15\.attention\.v_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.15\.attention\.v_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.15\.attention\.v_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.15\.attention\.v_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.15\.attention\.k_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.15\.attention\.k_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.15\.attention\.k_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.15\.attention\.k_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.15\.attention\.k_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.15\.attention\.k_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.15\.attention\.q_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.15\.attention\.q_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.15\.attention\.q_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.15\.attention\.q_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.15\.attention\.q_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.15\.attention\.q_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.15\.layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.15\.layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.15\.layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.15\.layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.15\.layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.15\.layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.14\.feed_forward\.output_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.14\.feed_forward\.output_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.14\.feed_forward\.output_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.14\.feed_forward\.output_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.14\.feed_forward\.output_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.14\.feed_forward\.output_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.14\.feed_forward\.intermediate_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.14\.feed_forward\.intermediate_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.14\.feed_forward\.intermediate_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.14\.feed_forward\.intermediate_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.14\.feed_forward\.intermediate_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.14\.feed_forward\.intermediate_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.14\.final_layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.14\.final_layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.14\.final_layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.14\.final_layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.14\.final_layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.14\.final_layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.14\.attention\.out_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.14\.attention\.out_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.14\.attention\.out_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.14\.attention\.out_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.14\.attention\.out_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.14\.attention\.out_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.14\.attention\.v_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.14\.attention\.v_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.14\.attention\.v_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.14\.attention\.v_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.14\.attention\.v_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.14\.attention\.v_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.14\.attention\.k_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.14\.attention\.k_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.14\.attention\.k_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.14\.attention\.k_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.14\.attention\.k_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.14\.attention\.k_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.14\.attention\.q_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.14\.attention\.q_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.14\.attention\.q_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.14\.attention\.q_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.14\.attention\.q_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.14\.attention\.q_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.14\.layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.14\.layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.14\.layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.14\.layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.14\.layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.14\.layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.13\.feed_forward\.output_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.13\.feed_forward\.output_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.13\.feed_forward\.output_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.13\.feed_forward\.output_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.13\.feed_forward\.output_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.13\.feed_forward\.output_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.13\.feed_forward\.intermediate_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.13\.feed_forward\.intermediate_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.13\.feed_forward\.intermediate_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.13\.feed_forward\.intermediate_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.13\.feed_forward\.intermediate_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.13\.feed_forward\.intermediate_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.13\.final_layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.13\.final_layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.13\.final_layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.13\.final_layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.13\.final_layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.13\.final_layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.13\.attention\.out_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.13\.attention\.out_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.13\.attention\.out_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.13\.attention\.out_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.13\.attention\.out_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.13\.attention\.out_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.13\.attention\.v_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.13\.attention\.v_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.13\.attention\.v_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.13\.attention\.v_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.13\.attention\.v_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.13\.attention\.v_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.13\.attention\.k_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.13\.attention\.k_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.13\.attention\.k_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.13\.attention\.k_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.13\.attention\.k_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.13\.attention\.k_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.13\.attention\.q_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.13\.attention\.q_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.13\.attention\.q_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.13\.attention\.q_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.13\.attention\.q_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.13\.attention\.q_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.13\.layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.13\.layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.13\.layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.13\.layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.13\.layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.13\.layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.37\.feed_forward\.output_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.37\.feed_forward\.output_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.37\.feed_forward\.output_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.37\.feed_forward\.output_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.37\.feed_forward\.output_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.37\.feed_forward\.output_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.37\.feed_forward\.intermediate_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.37\.feed_forward\.intermediate_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.37\.feed_forward\.intermediate_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.37\.feed_forward\.intermediate_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.37\.feed_forward\.intermediate_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.37\.feed_forward\.intermediate_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.37\.final_layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.37\.final_layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.37\.final_layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.37\.final_layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.37\.final_layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.37\.final_layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.37\.attention\.out_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.37\.attention\.out_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.37\.attention\.out_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.37\.attention\.out_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.37\.attention\.out_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.37\.attention\.out_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.37\.attention\.v_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.37\.attention\.v_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.37\.attention\.v_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.37\.attention\.v_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.37\.attention\.v_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.37\.attention\.v_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.37\.attention\.k_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.37\.attention\.k_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.37\.attention\.k_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.37\.attention\.k_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.37\.attention\.k_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.37\.attention\.k_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.37\.attention\.q_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.37\.attention\.q_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.37\.attention\.q_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.37\.attention\.q_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.37\.attention\.q_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.37\.attention\.q_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.37\.layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.37\.layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.37\.layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.37\.layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.37\.layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.37\.layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.36\.feed_forward\.output_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.36\.feed_forward\.output_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.36\.feed_forward\.output_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.36\.feed_forward\.output_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.36\.feed_forward\.output_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.36\.feed_forward\.output_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.36\.feed_forward\.intermediate_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.36\.feed_forward\.intermediate_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.36\.feed_forward\.intermediate_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.36\.feed_forward\.intermediate_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.36\.feed_forward\.intermediate_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.36\.feed_forward\.intermediate_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.36\.final_layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.36\.final_layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.36\.final_layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.36\.final_layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.36\.final_layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.36\.final_layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.36\.attention\.out_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.36\.attention\.out_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.36\.attention\.out_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.36\.attention\.out_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.36\.attention\.out_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.36\.attention\.out_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.36\.attention\.v_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.36\.attention\.v_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.36\.attention\.v_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.36\.attention\.v_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.36\.attention\.v_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.36\.attention\.v_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.36\.attention\.k_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.36\.attention\.k_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.36\.attention\.k_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.36\.attention\.k_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.36\.attention\.k_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.36\.attention\.k_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.36\.attention\.q_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.36\.attention\.q_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.36\.attention\.q_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.36\.attention\.q_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.36\.attention\.q_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.36\.attention\.q_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.36\.layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.36\.layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.36\.layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.36\.layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.36\.layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.36\.layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.35\.feed_forward\.output_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.35\.feed_forward\.output_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.35\.feed_forward\.output_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.35\.feed_forward\.output_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.35\.feed_forward\.output_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.35\.feed_forward\.output_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.35\.feed_forward\.intermediate_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.35\.feed_forward\.intermediate_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.35\.feed_forward\.intermediate_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.35\.feed_forward\.intermediate_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.35\.feed_forward\.intermediate_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.35\.feed_forward\.intermediate_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.35\.final_layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.35\.final_layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.35\.final_layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.35\.final_layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.35\.final_layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.35\.final_layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.35\.attention\.out_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.35\.attention\.out_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.35\.attention\.out_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.35\.attention\.out_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.35\.attention\.out_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.35\.attention\.out_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.35\.attention\.v_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.35\.attention\.v_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.35\.attention\.v_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.35\.attention\.v_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.35\.attention\.v_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.35\.attention\.v_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.35\.attention\.k_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.35\.attention\.k_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.35\.attention\.k_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.35\.attention\.k_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.35\.attention\.k_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.35\.attention\.k_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.35\.attention\.q_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.35\.attention\.q_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.35\.attention\.q_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.35\.attention\.q_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.35\.attention\.q_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.35\.attention\.q_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.35\.layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.35\.layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.35\.layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.35\.layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.35\.layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.35\.layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.25\.feed_forward\.output_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.25\.feed_forward\.output_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.25\.feed_forward\.output_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.25\.feed_forward\.output_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.25\.feed_forward\.output_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.25\.feed_forward\.output_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.25\.feed_forward\.intermediate_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.25\.feed_forward\.intermediate_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.25\.feed_forward\.intermediate_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.25\.feed_forward\.intermediate_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.25\.feed_forward\.intermediate_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.25\.feed_forward\.intermediate_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.25\.final_layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.25\.final_layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.25\.final_layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.25\.final_layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.25\.final_layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.25\.final_layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.25\.attention\.out_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.25\.attention\.out_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.25\.attention\.out_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.25\.attention\.out_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.25\.attention\.out_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.25\.attention\.out_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.25\.attention\.v_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.25\.attention\.v_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.25\.attention\.v_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.25\.attention\.v_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.25\.attention\.v_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.25\.attention\.v_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.25\.attention\.k_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.25\.attention\.k_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.25\.attention\.k_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.25\.attention\.k_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.25\.attention\.k_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.25\.attention\.k_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.25\.attention\.q_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.25\.attention\.q_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.25\.attention\.q_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.25\.attention\.q_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.25\.attention\.q_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.25\.attention\.q_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.25\.layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.25\.layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.25\.layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.25\.layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.25\.layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.25\.layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.22\.feed_forward\.output_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.22\.feed_forward\.output_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.22\.feed_forward\.output_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.22\.feed_forward\.output_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.22\.feed_forward\.output_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.22\.feed_forward\.output_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.22\.feed_forward\.intermediate_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.22\.feed_forward\.intermediate_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.22\.feed_forward\.intermediate_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.22\.feed_forward\.intermediate_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.22\.feed_forward\.intermediate_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.22\.feed_forward\.intermediate_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.22\.final_layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.22\.final_layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.22\.final_layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.22\.final_layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.22\.final_layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.22\.final_layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.22\.attention\.out_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.22\.attention\.out_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.22\.attention\.out_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.22\.attention\.out_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.22\.attention\.out_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.22\.attention\.out_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.22\.attention\.v_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.22\.attention\.v_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.22\.attention\.v_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.22\.attention\.v_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.22\.attention\.v_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.22\.attention\.v_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.22\.attention\.k_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.22\.attention\.k_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.22\.attention\.k_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.22\.attention\.k_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.22\.attention\.k_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.22\.attention\.k_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.22\.attention\.q_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.22\.attention\.q_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.22\.attention\.q_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.22\.attention\.q_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.22\.attention\.q_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.22\.attention\.q_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.22\.layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.22\.layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.22\.layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.22\.layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.22\.layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.22\.layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.20\.feed_forward\.output_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.20\.feed_forward\.output_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.20\.feed_forward\.output_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.20\.feed_forward\.output_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.20\.feed_forward\.output_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.20\.feed_forward\.output_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.20\.feed_forward\.intermediate_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.20\.feed_forward\.intermediate_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.20\.feed_forward\.intermediate_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.20\.feed_forward\.intermediate_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.20\.feed_forward\.intermediate_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.20\.feed_forward\.intermediate_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.20\.final_layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.20\.final_layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.20\.final_layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.20\.final_layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.20\.final_layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.20\.final_layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.20\.attention\.out_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.20\.attention\.out_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.20\.attention\.out_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.20\.attention\.out_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.20\.attention\.out_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.20\.attention\.out_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.20\.attention\.v_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.20\.attention\.v_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.20\.attention\.v_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.20\.attention\.v_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.20\.attention\.v_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.20\.attention\.v_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.20\.attention\.k_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.20\.attention\.k_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.20\.attention\.k_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.20\.attention\.k_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.20\.attention\.k_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.20\.attention\.k_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.20\.attention\.q_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.20\.attention\.q_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.20\.attention\.q_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.20\.attention\.q_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.20\.attention\.q_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.20\.attention\.q_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.20\.layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.20\.layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.20\.layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.20\.layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.20\.layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.20\.layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.12\.feed_forward\.output_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.12\.feed_forward\.output_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.12\.feed_forward\.output_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.12\.feed_forward\.output_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.12\.feed_forward\.output_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.12\.feed_forward\.output_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.12\.feed_forward\.intermediate_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.12\.feed_forward\.intermediate_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.12\.feed_forward\.intermediate_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.12\.feed_forward\.intermediate_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.12\.feed_forward\.intermediate_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.12\.feed_forward\.intermediate_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.12\.final_layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.12\.final_layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.12\.final_layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.12\.final_layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.12\.final_layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.12\.final_layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.12\.attention\.out_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.12\.attention\.out_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.12\.attention\.out_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.12\.attention\.out_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.12\.attention\.out_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.12\.attention\.out_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.12\.attention\.v_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.12\.attention\.v_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.12\.attention\.v_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.12\.attention\.v_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.12\.attention\.v_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.12\.attention\.v_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.12\.attention\.k_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.12\.attention\.k_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.12\.attention\.k_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.12\.attention\.k_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.12\.attention\.k_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.12\.attention\.k_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.12\.attention\.q_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.12\.attention\.q_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.12\.attention\.q_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.12\.attention\.q_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.12\.attention\.q_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.12\.attention\.q_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.12\.layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.12\.layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.12\.layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.12\.layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.12\.layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.12\.layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.10\.feed_forward\.output_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.10\.feed_forward\.output_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.10\.feed_forward\.output_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.10\.feed_forward\.output_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.10\.feed_forward\.output_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.10\.feed_forward\.output_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.10\.feed_forward\.intermediate_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.10\.feed_forward\.intermediate_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.10\.feed_forward\.intermediate_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.10\.feed_forward\.intermediate_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.10\.feed_forward\.intermediate_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.10\.feed_forward\.intermediate_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.10\.final_layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.10\.final_layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.10\.final_layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.10\.final_layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.10\.final_layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.10\.final_layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.10\.attention\.out_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.10\.attention\.out_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.10\.attention\.out_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.10\.attention\.out_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.10\.attention\.out_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.10\.attention\.out_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.10\.attention\.v_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.10\.attention\.v_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.10\.attention\.v_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.10\.attention\.v_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.10\.attention\.v_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.10\.attention\.v_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.10\.attention\.k_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.10\.attention\.k_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.10\.attention\.k_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.10\.attention\.k_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.10\.attention\.k_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.10\.attention\.k_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.10\.attention\.q_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.10\.attention\.q_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.10\.attention\.q_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.10\.attention\.q_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.10\.attention\.q_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.10\.attention\.q_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.10\.layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.10\.layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.10\.layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.10\.layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.10\.layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.10\.layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.6\.feed_forward\.output_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.6\.feed_forward\.output_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.6\.feed_forward\.output_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.6\.feed_forward\.output_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.6\.feed_forward\.output_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.6\.feed_forward\.output_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.6\.feed_forward\.intermediate_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.6\.feed_forward\.intermediate_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.6\.feed_forward\.intermediate_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.6\.feed_forward\.intermediate_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.6\.feed_forward\.intermediate_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.6\.feed_forward\.intermediate_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.6\.final_layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.6\.final_layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.6\.final_layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.6\.final_layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.6\.final_layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.6\.final_layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.6\.attention\.out_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.6\.attention\.out_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.6\.attention\.out_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.6\.attention\.out_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.6\.attention\.out_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.6\.attention\.out_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.6\.attention\.v_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.6\.attention\.v_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.6\.attention\.v_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.6\.attention\.v_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.6\.attention\.v_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.6\.attention\.v_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.6\.attention\.k_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.6\.attention\.k_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.6\.attention\.k_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.6\.attention\.k_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.6\.attention\.k_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.6\.attention\.k_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.6\.attention\.q_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.6\.attention\.q_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.6\.attention\.q_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.6\.attention\.q_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.6\.attention\.q_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.6\.attention\.q_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.6\.layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.6\.layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.6\.layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.6\.layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.6\.layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.6\.layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.4\.feed_forward\.output_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.4\.feed_forward\.output_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.4\.feed_forward\.output_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.4\.feed_forward\.output_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.4\.feed_forward\.output_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.4\.feed_forward\.output_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.4\.feed_forward\.intermediate_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.4\.feed_forward\.intermediate_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.4\.feed_forward\.intermediate_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.4\.feed_forward\.intermediate_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.4\.feed_forward\.intermediate_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.4\.feed_forward\.intermediate_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.4\.final_layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.4\.final_layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.4\.final_layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.4\.final_layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.4\.final_layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.4\.final_layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.4\.attention\.out_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.4\.attention\.out_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.4\.attention\.out_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.4\.attention\.out_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.4\.attention\.out_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.4\.attention\.out_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.4\.attention\.v_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.4\.attention\.v_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.4\.attention\.v_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.4\.attention\.v_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.4\.attention\.v_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.4\.attention\.v_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.4\.attention\.k_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.4\.attention\.k_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.4\.attention\.k_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.4\.attention\.k_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.4\.attention\.k_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.4\.attention\.k_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.4\.attention\.q_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.4\.attention\.q_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.4\.attention\.q_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.4\.attention\.q_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.4\.attention\.q_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.4\.attention\.q_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.4\.layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.4\.layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.4\.layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.4\.layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.4\.layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.4\.layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.0\.feed_forward\.output_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.0\.feed_forward\.output_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.0\.feed_forward\.output_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.0\.feed_forward\.output_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.0\.feed_forward\.output_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.0\.feed_forward\.output_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.0\.feed_forward\.intermediate_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.0\.feed_forward\.intermediate_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.0\.feed_forward\.intermediate_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.0\.feed_forward\.intermediate_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.0\.feed_forward\.intermediate_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.0\.feed_forward\.intermediate_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.0\.final_layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.0\.final_layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.0\.final_layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.0\.final_layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.0\.final_layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.0\.final_layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.0\.attention\.out_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.0\.attention\.out_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.0\.attention\.out_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.0\.attention\.out_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.0\.attention\.out_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.0\.attention\.out_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.0\.attention\.v_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.0\.attention\.v_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.0\.attention\.v_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.0\.attention\.v_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.0\.attention\.v_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.0\.attention\.v_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.0\.attention\.k_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.0\.attention\.k_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.0\.attention\.k_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.0\.attention\.k_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.0\.attention\.k_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.0\.attention\.k_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.0\.attention\.q_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.0\.attention\.q_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.0\.attention\.q_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.0\.attention\.q_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.0\.attention\.q_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.0\.attention\.q_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.0\.layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.0\.layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.0\.layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.0\.layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.0\.layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.0\.layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.47\.feed_forward\.output_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.47\.feed_forward\.output_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.47\.feed_forward\.output_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.47\.feed_forward\.output_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.47\.feed_forward\.output_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.47\.feed_forward\.output_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.47\.feed_forward\.intermediate_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.47\.feed_forward\.intermediate_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.47\.feed_forward\.intermediate_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.47\.feed_forward\.intermediate_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.47\.feed_forward\.intermediate_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.47\.feed_forward\.intermediate_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.47\.final_layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.47\.final_layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.47\.final_layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.47\.final_layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.47\.final_layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.47\.final_layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.47\.attention\.out_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.47\.attention\.out_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.47\.attention\.out_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.47\.attention\.out_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.47\.attention\.out_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.47\.attention\.out_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.47\.attention\.v_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.47\.attention\.v_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.47\.attention\.v_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.47\.attention\.v_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.47\.attention\.v_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.47\.attention\.v_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.47\.attention\.k_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.47\.attention\.k_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.47\.attention\.k_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.47\.attention\.k_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.47\.attention\.k_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.47\.attention\.k_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.47\.attention\.q_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.47\.attention\.q_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.47\.attention\.q_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.47\.attention\.q_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.47\.attention\.q_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.47\.attention\.q_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.47\.layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.47\.layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.47\.layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.47\.layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.47\.layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.47\.layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.44\.feed_forward\.output_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.44\.feed_forward\.output_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.44\.feed_forward\.output_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.44\.feed_forward\.output_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.44\.feed_forward\.output_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.44\.feed_forward\.output_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.44\.feed_forward\.intermediate_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.44\.feed_forward\.intermediate_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.44\.feed_forward\.intermediate_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.44\.feed_forward\.intermediate_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.44\.feed_forward\.intermediate_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.44\.feed_forward\.intermediate_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.44\.final_layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.44\.final_layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.44\.final_layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.44\.final_layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.44\.final_layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.44\.final_layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.44\.attention\.out_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.44\.attention\.out_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.44\.attention\.out_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.44\.attention\.out_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.44\.attention\.out_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.44\.attention\.out_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.44\.attention\.v_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.44\.attention\.v_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.44\.attention\.v_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.44\.attention\.v_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.44\.attention\.v_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.44\.attention\.v_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.44\.attention\.k_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.44\.attention\.k_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.44\.attention\.k_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.44\.attention\.k_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.44\.attention\.k_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.44\.attention\.k_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.44\.attention\.q_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.44\.attention\.q_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.44\.attention\.q_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.44\.attention\.q_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.44\.attention\.q_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.44\.attention\.q_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.44\.layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.44\.layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.44\.layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.44\.layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.44\.layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.44\.layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.40\.feed_forward\.output_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.40\.feed_forward\.output_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.40\.feed_forward\.output_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.40\.feed_forward\.output_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.40\.feed_forward\.output_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.40\.feed_forward\.output_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.40\.feed_forward\.intermediate_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.40\.feed_forward\.intermediate_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.40\.feed_forward\.intermediate_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.40\.feed_forward\.intermediate_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.40\.feed_forward\.intermediate_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.40\.feed_forward\.intermediate_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.40\.final_layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.40\.final_layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.40\.final_layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.40\.final_layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.40\.final_layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.40\.final_layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.40\.attention\.out_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.40\.attention\.out_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.40\.attention\.out_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.40\.attention\.out_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.40\.attention\.out_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.40\.attention\.out_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.40\.attention\.v_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.40\.attention\.v_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.40\.attention\.v_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.40\.attention\.v_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.40\.attention\.v_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.40\.attention\.v_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.40\.attention\.k_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.40\.attention\.k_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.40\.attention\.k_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.40\.attention\.k_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.40\.attention\.k_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.40\.attention\.k_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.40\.attention\.q_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.40\.attention\.q_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.40\.attention\.q_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.40\.attention\.q_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.40\.attention\.q_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.40\.attention\.q_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.40\.layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.40\.layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.40\.layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.40\.layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.40\.layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.40\.layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.39\.feed_forward\.output_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.39\.feed_forward\.output_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.39\.feed_forward\.output_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.39\.feed_forward\.output_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.39\.feed_forward\.output_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.39\.feed_forward\.output_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.39\.feed_forward\.intermediate_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.39\.feed_forward\.intermediate_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.39\.feed_forward\.intermediate_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.39\.feed_forward\.intermediate_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.39\.feed_forward\.intermediate_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.39\.feed_forward\.intermediate_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.39\.final_layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.39\.final_layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.39\.final_layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.39\.final_layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.39\.final_layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.39\.final_layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.39\.attention\.out_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.39\.attention\.out_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.39\.attention\.out_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.39\.attention\.out_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.39\.attention\.out_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.39\.attention\.out_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.39\.attention\.v_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.39\.attention\.v_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.39\.attention\.v_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.39\.attention\.v_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.39\.attention\.v_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.39\.attention\.v_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.39\.attention\.k_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.39\.attention\.k_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.39\.attention\.k_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.39\.attention\.k_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.39\.attention\.k_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.39\.attention\.k_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.39\.attention\.q_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.39\.attention\.q_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.39\.attention\.q_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.39\.attention\.q_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.39\.attention\.q_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.39\.attention\.q_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.39\.layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.39\.layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.39\.layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.39\.layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.39\.layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.39\.layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.38\.feed_forward\.output_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.38\.feed_forward\.output_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.38\.feed_forward\.output_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.38\.feed_forward\.output_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.38\.feed_forward\.output_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.38\.feed_forward\.output_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.38\.feed_forward\.intermediate_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.38\.feed_forward\.intermediate_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.38\.feed_forward\.intermediate_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.38\.feed_forward\.intermediate_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.38\.feed_forward\.intermediate_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.38\.feed_forward\.intermediate_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.38\.final_layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.38\.final_layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.38\.final_layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.38\.final_layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.38\.final_layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.38\.final_layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.38\.attention\.out_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.38\.attention\.out_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.38\.attention\.out_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.38\.attention\.out_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.38\.attention\.out_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.38\.attention\.out_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.38\.attention\.v_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.38\.attention\.v_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.38\.attention\.v_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.38\.attention\.v_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.38\.attention\.v_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.38\.attention\.v_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.38\.attention\.k_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.38\.attention\.k_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.38\.attention\.k_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.38\.attention\.k_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.38\.attention\.k_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.38\.attention\.k_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.38\.attention\.q_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.38\.attention\.q_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.38\.attention\.q_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.38\.attention\.q_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.38\.attention\.q_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.38\.attention\.q_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.38\.layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.38\.layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.38\.layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.38\.layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.38\.layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.38\.layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.34\.feed_forward\.output_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.34\.feed_forward\.output_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.34\.feed_forward\.output_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.34\.feed_forward\.output_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.34\.feed_forward\.output_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.34\.feed_forward\.output_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.34\.feed_forward\.intermediate_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.34\.feed_forward\.intermediate_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.34\.feed_forward\.intermediate_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.34\.feed_forward\.intermediate_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.34\.feed_forward\.intermediate_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.34\.feed_forward\.intermediate_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.34\.final_layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.34\.final_layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.34\.final_layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.34\.final_layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.34\.final_layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.34\.final_layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.34\.attention\.out_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.34\.attention\.out_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.34\.attention\.out_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.34\.attention\.out_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.34\.attention\.out_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.34\.attention\.out_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.34\.attention\.v_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.34\.attention\.v_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.34\.attention\.v_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.34\.attention\.v_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.34\.attention\.v_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.34\.attention\.v_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.34\.attention\.k_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.34\.attention\.k_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.34\.attention\.k_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.34\.attention\.k_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.34\.attention\.k_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.34\.attention\.k_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.34\.attention\.q_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.34\.attention\.q_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.34\.attention\.q_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.34\.attention\.q_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.34\.attention\.q_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.34\.attention\.q_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.34\.layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.34\.layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.34\.layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.34\.layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.34\.layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.34\.layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.31\.feed_forward\.output_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.31\.feed_forward\.output_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.31\.feed_forward\.output_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.31\.feed_forward\.output_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.31\.feed_forward\.output_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.31\.feed_forward\.output_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.31\.feed_forward\.intermediate_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.31\.feed_forward\.intermediate_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.31\.feed_forward\.intermediate_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.31\.feed_forward\.intermediate_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.31\.feed_forward\.intermediate_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.31\.feed_forward\.intermediate_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.31\.final_layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.31\.final_layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.31\.final_layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.31\.final_layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.31\.final_layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.31\.final_layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.31\.attention\.out_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.31\.attention\.out_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.31\.attention\.out_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.31\.attention\.out_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.31\.attention\.out_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.31\.attention\.out_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.31\.attention\.v_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.31\.attention\.v_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.31\.attention\.v_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.31\.attention\.v_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.31\.attention\.v_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.31\.attention\.v_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.31\.attention\.k_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.31\.attention\.k_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.31\.attention\.k_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.31\.attention\.k_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.31\.attention\.k_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.31\.attention\.k_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.31\.attention\.q_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.31\.attention\.q_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.31\.attention\.q_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.31\.attention\.q_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.31\.attention\.q_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.31\.attention\.q_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.31\.layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.31\.layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.31\.layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.31\.layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.31\.layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.31\.layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.28\.feed_forward\.output_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.28\.feed_forward\.output_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.28\.feed_forward\.output_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.28\.feed_forward\.output_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.28\.feed_forward\.output_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.28\.feed_forward\.output_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.28\.feed_forward\.intermediate_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.28\.feed_forward\.intermediate_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.28\.feed_forward\.intermediate_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.28\.feed_forward\.intermediate_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.28\.feed_forward\.intermediate_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.28\.feed_forward\.intermediate_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.28\.final_layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.28\.final_layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.28\.final_layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.28\.final_layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.28\.final_layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.28\.final_layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.28\.attention\.out_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.28\.attention\.out_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.28\.attention\.out_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.28\.attention\.out_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.28\.attention\.out_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.28\.attention\.out_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.28\.attention\.v_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.28\.attention\.v_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.28\.attention\.v_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.28\.attention\.v_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.28\.attention\.v_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.28\.attention\.v_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.28\.attention\.k_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.28\.attention\.k_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.28\.attention\.k_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.28\.attention\.k_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.28\.attention\.k_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.28\.attention\.k_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.28\.attention\.q_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.28\.attention\.q_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.28\.attention\.q_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.28\.attention\.q_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.28\.attention\.q_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.28\.attention\.q_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.28\.layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.28\.layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.28\.layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.28\.layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.28\.layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.28\.layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.27\.feed_forward\.output_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.27\.feed_forward\.output_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.27\.feed_forward\.output_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.27\.feed_forward\.output_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.27\.feed_forward\.output_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.27\.feed_forward\.output_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.27\.feed_forward\.intermediate_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.27\.feed_forward\.intermediate_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.27\.feed_forward\.intermediate_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.27\.feed_forward\.intermediate_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.27\.feed_forward\.intermediate_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.27\.feed_forward\.intermediate_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.27\.final_layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.27\.final_layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.27\.final_layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.27\.final_layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.27\.final_layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.27\.final_layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.27\.attention\.out_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.27\.attention\.out_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.27\.attention\.out_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.27\.attention\.out_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.27\.attention\.out_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.27\.attention\.out_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.27\.attention\.v_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.27\.attention\.v_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.27\.attention\.v_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.27\.attention\.v_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.27\.attention\.v_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.27\.attention\.v_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.27\.attention\.k_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.27\.attention\.k_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.27\.attention\.k_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.27\.attention\.k_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.27\.attention\.k_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.27\.attention\.k_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.27\.attention\.q_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.27\.attention\.q_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.27\.attention\.q_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.27\.attention\.q_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.27\.attention\.q_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.27\.attention\.q_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.27\.layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.27\.layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.27\.layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.27\.layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.27\.layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.27\.layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.26\.feed_forward\.output_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.26\.feed_forward\.output_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.26\.feed_forward\.output_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.26\.feed_forward\.output_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.26\.feed_forward\.output_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.26\.feed_forward\.output_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.26\.feed_forward\.intermediate_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.26\.feed_forward\.intermediate_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.26\.feed_forward\.intermediate_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.26\.feed_forward\.intermediate_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.26\.feed_forward\.intermediate_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.26\.feed_forward\.intermediate_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.26\.final_layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.26\.final_layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.26\.final_layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.26\.final_layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.26\.final_layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.26\.final_layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.26\.attention\.out_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.26\.attention\.out_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.26\.attention\.out_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.26\.attention\.out_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.26\.attention\.out_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.26\.attention\.out_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.26\.attention\.v_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.26\.attention\.v_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.26\.attention\.v_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.26\.attention\.v_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.26\.attention\.v_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.26\.attention\.v_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.26\.attention\.k_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.26\.attention\.k_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.26\.attention\.k_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.26\.attention\.k_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.26\.attention\.k_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.26\.attention\.k_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.26\.attention\.q_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.26\.attention\.q_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.26\.attention\.q_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.26\.attention\.q_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.26\.attention\.q_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.26\.attention\.q_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.26\.layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.26\.layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.26\.layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.26\.layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.26\.layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.26\.layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.21\.feed_forward\.output_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.21\.feed_forward\.output_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.21\.feed_forward\.output_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.21\.feed_forward\.output_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.21\.feed_forward\.output_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.21\.feed_forward\.output_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.21\.feed_forward\.intermediate_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.21\.feed_forward\.intermediate_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.21\.feed_forward\.intermediate_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.21\.feed_forward\.intermediate_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.21\.feed_forward\.intermediate_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.21\.feed_forward\.intermediate_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.21\.final_layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.21\.final_layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.21\.final_layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.21\.final_layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.21\.final_layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.21\.final_layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.21\.attention\.out_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.21\.attention\.out_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.21\.attention\.out_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.21\.attention\.out_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.21\.attention\.out_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.21\.attention\.out_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.21\.attention\.v_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.21\.attention\.v_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.21\.attention\.v_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.21\.attention\.v_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.21\.attention\.v_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.21\.attention\.v_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.21\.attention\.k_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.21\.attention\.k_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.21\.attention\.k_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.21\.attention\.k_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.21\.attention\.k_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.21\.attention\.k_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.21\.attention\.q_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.21\.attention\.q_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.21\.attention\.q_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.21\.attention\.q_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.21\.attention\.q_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.21\.attention\.q_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.21\.layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.21\.layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.21\.layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.21\.layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.21\.layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.21\.layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.8\.feed_forward\.output_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.8\.feed_forward\.output_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.8\.feed_forward\.output_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.8\.feed_forward\.output_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.8\.feed_forward\.output_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.8\.feed_forward\.output_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.8\.feed_forward\.intermediate_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.8\.feed_forward\.intermediate_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.8\.feed_forward\.intermediate_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.8\.feed_forward\.intermediate_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.8\.feed_forward\.intermediate_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.8\.feed_forward\.intermediate_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.8\.final_layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.8\.final_layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.8\.final_layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.8\.final_layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.8\.final_layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.8\.final_layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.8\.attention\.out_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.8\.attention\.out_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.8\.attention\.out_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.8\.attention\.out_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.8\.attention\.out_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.8\.attention\.out_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.8\.attention\.v_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.8\.attention\.v_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.8\.attention\.v_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.8\.attention\.v_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.8\.attention\.v_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.8\.attention\.v_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.8\.attention\.k_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.8\.attention\.k_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.8\.attention\.k_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.8\.attention\.k_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.8\.attention\.k_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.8\.attention\.k_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.8\.attention\.q_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.8\.attention\.q_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.8\.attention\.q_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.8\.attention\.q_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.8\.attention\.q_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.8\.attention\.q_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.8\.layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.8\.layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.8\.layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.8\.layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.8\.layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.8\.layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.45\.feed_forward\.output_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.45\.feed_forward\.output_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.45\.feed_forward\.output_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.45\.feed_forward\.output_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.45\.feed_forward\.output_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.45\.feed_forward\.output_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.45\.feed_forward\.intermediate_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.45\.feed_forward\.intermediate_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.45\.feed_forward\.intermediate_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.45\.feed_forward\.intermediate_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.45\.feed_forward\.intermediate_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.45\.feed_forward\.intermediate_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.45\.final_layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.45\.final_layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.45\.final_layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.45\.final_layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.45\.final_layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.45\.final_layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.45\.attention\.out_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.45\.attention\.out_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.45\.attention\.out_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.45\.attention\.out_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.45\.attention\.out_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.45\.attention\.out_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.45\.attention\.v_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.45\.attention\.v_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.45\.attention\.v_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.45\.attention\.v_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.45\.attention\.v_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.45\.attention\.v_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.45\.attention\.k_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.45\.attention\.k_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.45\.attention\.k_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.45\.attention\.k_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.45\.attention\.k_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.45\.attention\.k_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.45\.attention\.q_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.45\.attention\.q_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.45\.attention\.q_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.45\.attention\.q_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.45\.attention\.q_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.45\.attention\.q_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.45\.layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.45\.layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.45\.layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.45\.layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.45\.layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.45\.layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.43\.feed_forward\.output_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.43\.feed_forward\.output_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.43\.feed_forward\.output_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.43\.feed_forward\.output_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.43\.feed_forward\.output_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.43\.feed_forward\.output_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.43\.feed_forward\.intermediate_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.43\.feed_forward\.intermediate_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.43\.feed_forward\.intermediate_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.43\.feed_forward\.intermediate_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.43\.feed_forward\.intermediate_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.43\.feed_forward\.intermediate_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.43\.final_layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.43\.final_layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.43\.final_layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.43\.final_layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.43\.final_layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.43\.final_layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.43\.attention\.out_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.43\.attention\.out_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.43\.attention\.out_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.43\.attention\.out_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.43\.attention\.out_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.43\.attention\.out_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.43\.attention\.v_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.43\.attention\.v_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.43\.attention\.v_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.43\.attention\.v_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.43\.attention\.v_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.43\.attention\.v_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.43\.attention\.k_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.43\.attention\.k_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.43\.attention\.k_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.43\.attention\.k_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.43\.attention\.k_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.43\.attention\.k_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.43\.attention\.q_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.43\.attention\.q_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.43\.attention\.q_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.43\.attention\.q_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.43\.attention\.q_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.43\.attention\.q_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.43\.layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.43\.layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.43\.layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.43\.layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.43\.layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.43\.layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.29\.feed_forward\.output_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.29\.feed_forward\.output_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.29\.feed_forward\.output_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.29\.feed_forward\.output_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.29\.feed_forward\.output_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.29\.feed_forward\.output_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.29\.feed_forward\.intermediate_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.29\.feed_forward\.intermediate_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.29\.feed_forward\.intermediate_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.29\.feed_forward\.intermediate_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.29\.feed_forward\.intermediate_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.29\.feed_forward\.intermediate_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.29\.final_layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.29\.final_layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.29\.final_layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.29\.final_layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.29\.final_layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.29\.final_layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.29\.attention\.out_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.29\.attention\.out_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.29\.attention\.out_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.29\.attention\.out_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.29\.attention\.out_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.29\.attention\.out_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.29\.attention\.v_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.29\.attention\.v_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.29\.attention\.v_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.29\.attention\.v_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.29\.attention\.v_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.29\.attention\.v_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.29\.attention\.k_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.29\.attention\.k_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.29\.attention\.k_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.29\.attention\.k_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.29\.attention\.k_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.29\.attention\.k_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.29\.attention\.q_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.29\.attention\.q_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.29\.attention\.q_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.29\.attention\.q_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.29\.attention\.q_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.29\.attention\.q_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.29\.layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.29\.layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.29\.layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.29\.layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.29\.layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.29\.layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.19\.feed_forward\.output_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.19\.feed_forward\.output_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.19\.feed_forward\.output_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.19\.feed_forward\.output_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.19\.feed_forward\.output_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.19\.feed_forward\.output_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.19\.feed_forward\.intermediate_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.19\.feed_forward\.intermediate_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.19\.feed_forward\.intermediate_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.19\.feed_forward\.intermediate_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.19\.feed_forward\.intermediate_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.19\.feed_forward\.intermediate_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.19\.final_layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.19\.final_layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.19\.final_layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.19\.final_layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.19\.final_layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.19\.final_layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.19\.attention\.out_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.19\.attention\.out_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.19\.attention\.out_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.19\.attention\.out_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.19\.attention\.out_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.19\.attention\.out_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.19\.attention\.v_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.19\.attention\.v_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.19\.attention\.v_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.19\.attention\.v_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.19\.attention\.v_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.19\.attention\.v_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.19\.attention\.k_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.19\.attention\.k_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.19\.attention\.k_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.19\.attention\.k_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.19\.attention\.k_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.19\.attention\.k_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.19\.attention\.q_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.19\.attention\.q_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.19\.attention\.q_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.19\.attention\.q_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.19\.attention\.q_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.19\.attention\.q_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.19\.layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.19\.layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.19\.layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.19\.layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.19\.layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.19\.layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.18\.feed_forward\.output_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.18\.feed_forward\.output_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.18\.feed_forward\.output_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.18\.feed_forward\.output_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.18\.feed_forward\.output_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.18\.feed_forward\.output_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.18\.feed_forward\.intermediate_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.18\.feed_forward\.intermediate_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.18\.feed_forward\.intermediate_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.18\.feed_forward\.intermediate_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.18\.feed_forward\.intermediate_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.18\.feed_forward\.intermediate_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.18\.final_layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.18\.final_layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.18\.final_layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.18\.final_layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.18\.final_layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.18\.final_layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.18\.attention\.out_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.18\.attention\.out_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.18\.attention\.out_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.18\.attention\.out_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.18\.attention\.out_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.18\.attention\.out_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.18\.attention\.v_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.18\.attention\.v_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.18\.attention\.v_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.18\.attention\.v_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.18\.attention\.v_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.18\.attention\.v_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.18\.attention\.k_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.18\.attention\.k_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.18\.attention\.k_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.18\.attention\.k_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.18\.attention\.k_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.18\.attention\.k_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.18\.attention\.q_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.18\.attention\.q_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.18\.attention\.q_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.18\.attention\.q_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.18\.attention\.q_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.18\.attention\.q_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.18\.layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.18\.layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.18\.layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.18\.layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.18\.layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.18\.layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.17\.feed_forward\.output_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.17\.feed_forward\.output_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.17\.feed_forward\.output_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.17\.feed_forward\.output_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.17\.feed_forward\.output_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.17\.feed_forward\.output_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.17\.feed_forward\.intermediate_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.17\.feed_forward\.intermediate_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.17\.feed_forward\.intermediate_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.17\.feed_forward\.intermediate_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.17\.feed_forward\.intermediate_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.17\.feed_forward\.intermediate_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.17\.final_layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.17\.final_layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.17\.final_layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.17\.final_layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.17\.final_layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.17\.final_layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.17\.attention\.out_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.17\.attention\.out_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.17\.attention\.out_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.17\.attention\.out_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.17\.attention\.out_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.17\.attention\.out_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.17\.attention\.v_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.17\.attention\.v_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.17\.attention\.v_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.17\.attention\.v_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.17\.attention\.v_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.17\.attention\.v_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.17\.attention\.k_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.17\.attention\.k_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.17\.attention\.k_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.17\.attention\.k_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.17\.attention\.k_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.17\.attention\.k_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.17\.attention\.q_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.17\.attention\.q_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.17\.attention\.q_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.17\.attention\.q_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.17\.attention\.q_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.17\.attention\.q_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.17\.layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.17\.layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.17\.layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.17\.layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.17\.layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.17\.layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.16\.feed_forward\.output_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.16\.feed_forward\.output_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.16\.feed_forward\.output_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.16\.feed_forward\.output_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.16\.feed_forward\.output_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.16\.feed_forward\.output_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.16\.feed_forward\.intermediate_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.16\.feed_forward\.intermediate_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.16\.feed_forward\.intermediate_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.16\.feed_forward\.intermediate_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.16\.feed_forward\.intermediate_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.16\.feed_forward\.intermediate_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.16\.final_layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.16\.final_layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.16\.final_layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.16\.final_layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.16\.final_layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.16\.final_layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.16\.attention\.out_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.16\.attention\.out_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.16\.attention\.out_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.16\.attention\.out_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.16\.attention\.out_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.16\.attention\.out_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.16\.attention\.v_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.16\.attention\.v_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.16\.attention\.v_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.16\.attention\.v_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.16\.attention\.v_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.16\.attention\.v_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.16\.attention\.k_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.16\.attention\.k_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.16\.attention\.k_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.16\.attention\.k_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.16\.attention\.k_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.16\.attention\.k_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.16\.attention\.q_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.16\.attention\.q_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.16\.attention\.q_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.16\.attention\.q_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.16\.attention\.q_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.16\.attention\.q_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.16\.layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.16\.layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.16\.layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.16\.layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.16\.layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.16\.layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.9\.feed_forward\.output_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.9\.feed_forward\.output_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.9\.feed_forward\.output_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.9\.feed_forward\.output_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.9\.feed_forward\.output_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.9\.feed_forward\.output_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.9\.feed_forward\.intermediate_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.9\.feed_forward\.intermediate_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.9\.feed_forward\.intermediate_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.9\.feed_forward\.intermediate_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.9\.feed_forward\.intermediate_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.9\.feed_forward\.intermediate_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.9\.final_layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.9\.final_layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.9\.final_layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.9\.final_layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.9\.final_layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.9\.final_layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.9\.attention\.out_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.9\.attention\.out_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.9\.attention\.out_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.9\.attention\.out_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.9\.attention\.out_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.9\.attention\.out_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.9\.attention\.v_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.9\.attention\.v_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.9\.attention\.v_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.9\.attention\.v_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.9\.attention\.v_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.9\.attention\.v_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.9\.attention\.k_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.9\.attention\.k_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.9\.attention\.k_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.9\.attention\.k_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.9\.attention\.k_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.9\.attention\.k_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.9\.attention\.q_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.9\.attention\.q_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.9\.attention\.q_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.9\.attention\.q_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.9\.attention\.q_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.9\.attention\.q_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.9\.layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.9\.layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.9\.layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.9\.layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.9\.layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.9\.layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.7\.feed_forward\.output_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.7\.feed_forward\.output_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.7\.feed_forward\.output_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.7\.feed_forward\.output_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.7\.feed_forward\.output_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.7\.feed_forward\.output_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.7\.feed_forward\.intermediate_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.7\.feed_forward\.intermediate_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.7\.feed_forward\.intermediate_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.7\.feed_forward\.intermediate_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.7\.feed_forward\.intermediate_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.7\.feed_forward\.intermediate_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.7\.final_layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.7\.final_layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.7\.final_layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.7\.final_layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.7\.final_layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.7\.final_layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.7\.attention\.out_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.7\.attention\.out_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.7\.attention\.out_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.7\.attention\.out_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.7\.attention\.out_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.7\.attention\.out_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.7\.attention\.v_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.7\.attention\.v_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.7\.attention\.v_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.7\.attention\.v_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.7\.attention\.v_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.7\.attention\.v_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.7\.attention\.k_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.7\.attention\.k_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.7\.attention\.k_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.7\.attention\.k_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.7\.attention\.k_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.7\.attention\.k_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.7\.attention\.q_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.7\.attention\.q_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.7\.attention\.q_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.7\.attention\.q_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.7\.attention\.q_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.7\.attention\.q_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.7\.layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.7\.layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.7\.layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.7\.layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.7\.layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.7\.layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.5\.feed_forward\.output_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.5\.feed_forward\.output_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.5\.feed_forward\.output_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.5\.feed_forward\.output_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.5\.feed_forward\.output_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.5\.feed_forward\.output_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.5\.feed_forward\.intermediate_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.5\.feed_forward\.intermediate_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.5\.feed_forward\.intermediate_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.5\.feed_forward\.intermediate_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.5\.feed_forward\.intermediate_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.5\.feed_forward\.intermediate_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.5\.final_layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.5\.final_layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.5\.final_layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.5\.final_layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.5\.final_layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.5\.final_layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.5\.attention\.out_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.5\.attention\.out_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.5\.attention\.out_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.5\.attention\.out_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.5\.attention\.out_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.5\.attention\.out_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.5\.attention\.v_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.5\.attention\.v_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.5\.attention\.v_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.5\.attention\.v_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.5\.attention\.v_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.5\.attention\.v_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.5\.attention\.k_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.5\.attention\.k_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.5\.attention\.k_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.5\.attention\.k_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.5\.attention\.k_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.5\.attention\.k_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.5\.attention\.q_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.5\.attention\.q_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.5\.attention\.q_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.5\.attention\.q_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.5\.attention\.q_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.5\.attention\.q_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.5\.layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.5\.layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.5\.layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.5\.layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.5\.layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.5\.layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.2\.feed_forward\.output_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.2\.feed_forward\.output_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.2\.feed_forward\.output_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.2\.feed_forward\.output_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.2\.feed_forward\.output_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.2\.feed_forward\.output_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.2\.feed_forward\.intermediate_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.2\.feed_forward\.intermediate_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.2\.feed_forward\.intermediate_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.2\.feed_forward\.intermediate_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.2\.feed_forward\.intermediate_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.2\.feed_forward\.intermediate_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.2\.final_layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.2\.final_layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.2\.final_layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.2\.final_layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.2\.final_layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.2\.final_layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.2\.attention\.out_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.2\.attention\.out_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.2\.attention\.out_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.2\.attention\.out_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.2\.attention\.out_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.2\.attention\.out_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.2\.attention\.v_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.2\.attention\.v_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.2\.attention\.v_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.2\.attention\.v_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.2\.attention\.v_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.2\.attention\.v_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.2\.attention\.k_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.2\.attention\.k_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.2\.attention\.k_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.2\.attention\.k_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.2\.attention\.k_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.2\.attention\.k_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.2\.attention\.q_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.2\.attention\.q_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.2\.attention\.q_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.2\.attention\.q_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.2\.attention\.q_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.2\.attention\.q_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.2\.layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.2\.layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.2\.layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.2\.layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.2\.layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.2\.layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.1\.feed_forward\.output_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.1\.feed_forward\.output_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.1\.feed_forward\.output_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.1\.feed_forward\.output_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.1\.feed_forward\.output_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.1\.feed_forward\.output_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.1\.feed_forward\.intermediate_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.1\.feed_forward\.intermediate_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.1\.feed_forward\.intermediate_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.1\.feed_forward\.intermediate_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.1\.feed_forward\.intermediate_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.1\.feed_forward\.intermediate_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.1\.final_layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.1\.final_layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.1\.final_layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.1\.final_layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.1\.final_layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.1\.final_layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.1\.attention\.out_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.1\.attention\.out_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.1\.attention\.out_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.1\.attention\.out_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.1\.attention\.out_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.1\.attention\.out_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.1\.attention\.v_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.1\.attention\.v_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.1\.attention\.v_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.1\.attention\.v_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.1\.attention\.v_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.1\.attention\.v_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.1\.attention\.k_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.1\.attention\.k_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.1\.attention\.k_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.1\.attention\.k_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.1\.attention\.k_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.1\.attention\.k_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.1\.attention\.q_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.1\.attention\.q_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.1\.attention\.q_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.1\.attention\.q_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.1\.attention\.q_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.1\.attention\.q_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.1\.layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.1\.layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.1\.layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.1\.layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.1\.layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.1\.layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.41\.feed_forward\.output_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.41\.feed_forward\.output_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.41\.feed_forward\.output_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.41\.feed_forward\.output_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.41\.feed_forward\.output_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.41\.feed_forward\.output_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.41\.feed_forward\.intermediate_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.41\.feed_forward\.intermediate_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.41\.feed_forward\.intermediate_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.41\.feed_forward\.intermediate_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.41\.feed_forward\.intermediate_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.41\.feed_forward\.intermediate_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.41\.final_layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.41\.final_layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.41\.final_layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.41\.final_layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.41\.final_layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.41\.final_layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.41\.attention\.out_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.41\.attention\.out_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.41\.attention\.out_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.41\.attention\.out_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.41\.attention\.out_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.41\.attention\.out_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.41\.attention\.v_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.41\.attention\.v_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.41\.attention\.v_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.41\.attention\.v_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.41\.attention\.v_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.41\.attention\.v_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.41\.attention\.k_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.41\.attention\.k_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.41\.attention\.k_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.41\.attention\.k_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.41\.attention\.k_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.41\.attention\.k_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.41\.attention\.q_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.41\.attention\.q_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.41\.attention\.q_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.41\.attention\.q_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.41\.attention\.q_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.41\.attention\.q_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.41\.layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.41\.layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.41\.layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.41\.layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.41\.layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.41\.layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.33\.feed_forward\.output_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.33\.feed_forward\.output_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.33\.feed_forward\.output_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.33\.feed_forward\.output_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.33\.feed_forward\.output_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.33\.feed_forward\.output_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.33\.feed_forward\.intermediate_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.33\.feed_forward\.intermediate_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.33\.feed_forward\.intermediate_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.33\.feed_forward\.intermediate_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.33\.feed_forward\.intermediate_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.33\.feed_forward\.intermediate_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.33\.final_layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.33\.final_layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.33\.final_layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.33\.final_layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.33\.final_layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.33\.final_layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.33\.attention\.out_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.33\.attention\.out_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.33\.attention\.out_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.33\.attention\.out_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.33\.attention\.out_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.33\.attention\.out_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.33\.attention\.v_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.33\.attention\.v_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.33\.attention\.v_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.33\.attention\.v_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.33\.attention\.v_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.33\.attention\.v_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.33\.attention\.k_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.33\.attention\.k_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.33\.attention\.k_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.33\.attention\.k_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.33\.attention\.k_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.33\.attention\.k_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.33\.attention\.q_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.33\.attention\.q_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.33\.attention\.q_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.33\.attention\.q_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.33\.attention\.q_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.33\.attention\.q_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.33\.layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.33\.layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.33\.layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.33\.layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.33\.layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.33\.layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.30\.feed_forward\.output_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.30\.feed_forward\.output_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.30\.feed_forward\.output_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.30\.feed_forward\.output_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.30\.feed_forward\.output_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.30\.feed_forward\.output_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.30\.feed_forward\.intermediate_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.30\.feed_forward\.intermediate_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.30\.feed_forward\.intermediate_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.30\.feed_forward\.intermediate_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.30\.feed_forward\.intermediate_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.30\.feed_forward\.intermediate_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.30\.final_layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.30\.final_layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.30\.final_layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.30\.final_layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.30\.final_layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.30\.final_layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.30\.attention\.out_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.30\.attention\.out_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.30\.attention\.out_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.30\.attention\.out_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.30\.attention\.out_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.30\.attention\.out_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.30\.attention\.v_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.30\.attention\.v_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.30\.attention\.v_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.30\.attention\.v_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.30\.attention\.v_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.30\.attention\.v_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.30\.attention\.k_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.30\.attention\.k_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.30\.attention\.k_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.30\.attention\.k_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.30\.attention\.k_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.30\.attention\.k_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.30\.attention\.q_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.30\.attention\.q_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.30\.attention\.q_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.30\.attention\.q_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.30\.attention\.q_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.30\.attention\.q_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.30\.layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.30\.layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.30\.layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.30\.layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.30\.layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.30\.layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.24\.feed_forward\.output_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.24\.feed_forward\.output_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.24\.feed_forward\.output_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.24\.feed_forward\.output_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.24\.feed_forward\.output_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.24\.feed_forward\.output_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.24\.feed_forward\.intermediate_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.24\.feed_forward\.intermediate_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.24\.feed_forward\.intermediate_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.24\.feed_forward\.intermediate_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.24\.feed_forward\.intermediate_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.24\.feed_forward\.intermediate_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.24\.final_layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.24\.final_layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.24\.final_layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.24\.final_layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.24\.final_layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.24\.final_layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.24\.attention\.out_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.24\.attention\.out_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.24\.attention\.out_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.24\.attention\.out_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.24\.attention\.out_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.24\.attention\.out_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.24\.attention\.v_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.24\.attention\.v_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.24\.attention\.v_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.24\.attention\.v_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.24\.attention\.v_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.24\.attention\.v_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.24\.attention\.k_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.24\.attention\.k_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.24\.attention\.k_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.24\.attention\.k_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.24\.attention\.k_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.24\.attention\.k_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.24\.attention\.q_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.24\.attention\.q_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.24\.attention\.q_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.24\.attention\.q_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.24\.attention\.q_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.24\.attention\.q_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.24\.layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.24\.layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.24\.layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.24\.layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.24\.layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.24\.layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.32\.feed_forward\.output_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.32\.feed_forward\.output_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.32\.feed_forward\.output_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.32\.feed_forward\.output_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.32\.feed_forward\.output_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.32\.feed_forward\.output_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.32\.feed_forward\.intermediate_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.32\.feed_forward\.intermediate_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.32\.feed_forward\.intermediate_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.32\.feed_forward\.intermediate_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.32\.feed_forward\.intermediate_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.32\.feed_forward\.intermediate_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.32\.final_layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.32\.final_layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.32\.final_layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.32\.final_layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.32\.final_layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.32\.final_layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.32\.attention\.out_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.32\.attention\.out_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.32\.attention\.out_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.32\.attention\.out_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.32\.attention\.out_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.32\.attention\.out_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.32\.attention\.v_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.32\.attention\.v_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.32\.attention\.v_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.32\.attention\.v_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.32\.attention\.v_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.32\.attention\.v_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.32\.attention\.k_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.32\.attention\.k_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.32\.attention\.k_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.32\.attention\.k_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.32\.attention\.k_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.32\.attention\.k_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.32\.attention\.q_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.32\.attention\.q_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.32\.attention\.q_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.32\.attention\.q_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.32\.attention\.q_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.32\.attention\.q_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.32\.layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.32\.layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.32\.layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.32\.layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.32\.layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.32\.layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.23\.feed_forward\.output_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.23\.feed_forward\.output_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.23\.feed_forward\.output_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.23\.feed_forward\.output_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.23\.feed_forward\.output_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.23\.feed_forward\.output_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.23\.feed_forward\.intermediate_dense\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.23\.feed_forward\.intermediate_dense\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.23\.feed_forward\.intermediate_dense\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.23\.feed_forward\.intermediate_dense\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.23\.feed_forward\.intermediate_dense\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.23\.feed_forward\.intermediate_dense\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.23\.final_layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.23\.final_layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.23\.final_layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.23\.final_layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.23\.final_layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.23\.final_layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.23\.attention\.out_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.23\.attention\.out_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.23\.attention\.out_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.23\.attention\.out_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.23\.attention\.out_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.23\.attention\.out_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.23\.attention\.v_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.23\.attention\.v_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.23\.attention\.v_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.23\.attention\.v_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.23\.attention\.v_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.23\.attention\.v_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.23\.attention\.k_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.23\.attention\.k_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.23\.attention\.k_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.23\.attention\.k_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.23\.attention\.k_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.23\.attention\.k_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.23\.attention\.q_proj\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.23\.attention\.q_proj\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.23\.attention\.q_proj\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.23\.attention\.q_proj\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.23\.attention\.q_proj\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.23\.attention\.q_proj\.bias.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.23\.layer_norm\.weight._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.23\.layer_norm\.weight.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.23\.layer_norm\.weight.bins + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.23\.layer_norm\.bias._type + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.23\.layer_norm\.bias.values + 5: 1 + 6: + - 1 + - 1: gradients/wav2vec2\.encoder\.layers\.23\.layer_norm\.bias.bins + 5: 1 + 6: + - 1 + - 1: train/loss + 5: 1 + 6: + - 1 + - 1: train/learning_rate + 5: 1 + 6: + - 1 + - 1: train/epoch + 5: 1 + 6: + - 1 + - 1: eval/loss + 5: 1 + 6: + - 1 + - 1: eval/wer + 5: 1 + 6: + - 1 + - 1: eval/runtime + 5: 1 + 6: + - 1 + - 1: eval/samples_per_second + 5: 1 + 6: + - 1 + - 1: eval/steps_per_second + 5: 1 + 6: + - 1 + python_version: 3.9.7 + start_time: 1655371226 + t: + 1: + - 1 + - 5 + - 11 + - 49 + - 51 + - 53 + - 55 + 3: + - 1 + - 7 + - 13 + 4: 3.9.7 + 5: 0.12.15 + 6: 4.18.0 + 8: + - 5 +activation_dropout: + desc: null + value: 0.055 +adafactor: + desc: null + value: false +adam_beta1: + desc: null + value: 0.9 +adam_beta2: + desc: null + value: 0.999 +adam_epsilon: + desc: null + value: 1.0e-08 +adapter_kernel_size: + desc: null + value: 3 +adapter_stride: + desc: null + value: 2 +add_adapter: + desc: null + value: false +add_cross_attention: + desc: null + value: false +apply_spec_augment: + desc: null + value: true +architectures: + desc: null + value: + - Wav2Vec2ForPreTraining +attention_dropout: + desc: null + value: 0.094 +bad_words_ids: + desc: null + value: null +bf16: + desc: null + value: false +bf16_full_eval: + desc: null + value: false +bos_token_id: + desc: null + value: 1 +chunk_size_feed_forward: + desc: null + value: 0 +classifier_proj_size: + desc: null + value: 256 +codevector_dim: + desc: null + value: 1024 +contrastive_logits_temperature: + desc: null + value: 0.1 +conv_bias: + desc: null + value: true +conv_dim: + desc: null + value: + - 512 + - 512 + - 512 + - 512 + - 512 + - 512 + - 512 +conv_kernel: + desc: null + value: + - 10 + - 3 + - 3 + - 3 + - 3 + - 2 + - 2 +conv_stride: + desc: null + value: + - 5 + - 2 + - 2 + - 2 + - 2 + - 2 + - 2 +cross_attention_hidden_size: + desc: null + value: null +ctc_loss_reduction: + desc: null + value: mean +ctc_zero_infinity: + desc: null + value: true +data_seed: + desc: null + value: None +dataloader_drop_last: + desc: null + value: false +dataloader_num_workers: + desc: null + value: 0 +dataloader_pin_memory: + desc: null + value: true +ddp_bucket_cap_mb: + desc: null + value: None +ddp_find_unused_parameters: + desc: null + value: None +debug: + desc: null + value: '[]' +decoder_start_token_id: + desc: null + value: null +deepspeed: + desc: null + value: None +disable_tqdm: + desc: null + value: false +diversity_loss_weight: + desc: null + value: 0.1 +diversity_penalty: + desc: null + value: 0.0 +do_eval: + desc: null + value: true +do_predict: + desc: null + value: false +do_sample: + desc: null + value: false +do_stable_layer_norm: + desc: null + value: true +do_train: + desc: null + value: true +early_stopping: + desc: null + value: false +encoder_no_repeat_ngram_size: + desc: null + value: 0 +eos_token_id: + desc: null + value: 2 +eval_accumulation_steps: + desc: null + value: None +eval_batch_size: + desc: null + value: 12 +eval_delay: + desc: null + value: 0 +eval_steps: + desc: null + value: 500 +evaluation_strategy: + desc: null + value: steps +exponential_decay_length_penalty: + desc: null + value: null +feat_extract_activation: + desc: null + value: gelu +feat_extract_dropout: + desc: null + value: 0.0 +feat_extract_norm: + desc: null + value: layer +feat_proj_dropout: + desc: null + value: 0.04 +feat_quantizer_dropout: + desc: null + value: 0.0 +final_dropout: + desc: null + value: 0.0 +finetuning_task: + desc: null + value: null +forced_bos_token_id: + desc: null + value: null +forced_eos_token_id: + desc: null + value: null +fp16: + desc: null + value: true +fp16_backend: + desc: null + value: auto +fp16_full_eval: + desc: null + value: false +fp16_opt_level: + desc: null + value: O1 +gradient_accumulation_steps: + desc: null + value: 2 +gradient_checkpointing: + desc: null + value: true +greater_is_better: + desc: null + value: None +group_by_length: + desc: null + value: true +half_precision_backend: + desc: null + value: amp +hidden_act: + desc: null + value: gelu +hidden_dropout: + desc: null + value: 0.047 +hidden_size: + desc: null + value: 1280 +hub_model_id: + desc: null + value: NbAiLab/wav2vec2-1b-npsc-nst-bokmaal +hub_strategy: + desc: null + value: every_save +hub_token: + desc: null + value: +id2label: + desc: null + value: + '0': LABEL_0 + '1': LABEL_1 +ignore_data_skip: + desc: null + value: false +initializer_range: + desc: null + value: 0.02 +intermediate_size: + desc: null + value: 5120 +is_decoder: + desc: null + value: false +is_encoder_decoder: + desc: null + value: false +label2id: + desc: null + value: + LABEL_0: 0 + LABEL_1: 1 +label_names: + desc: null + value: None +label_smoothing_factor: + desc: null + value: 0.0 +layer_norm_eps: + desc: null + value: 1.0e-05 +layerdrop: + desc: null + value: 0.041 +learning_rate: + desc: null + value: 2.0e-05 +length_column_name: + desc: null + value: input_length +length_penalty: + desc: null + value: 1.0 +load_best_model_at_end: + desc: null + value: false +local_rank: + desc: null + value: -1 +log_level: + desc: null + value: -1 +log_level_replica: + desc: null + value: -1 +log_on_each_node: + desc: null + value: true +logging_dir: + desc: null + value: ./runs/Jun16_11-19-30_dante +logging_first_step: + desc: null + value: false +logging_nan_inf_filter: + desc: null + value: true +logging_steps: + desc: null + value: 100 +logging_strategy: + desc: null + value: steps +lr_scheduler_type: + desc: null + value: linear +mask_feature_length: + desc: null + value: 64 +mask_feature_min_masks: + desc: null + value: 0 +mask_feature_prob: + desc: null + value: 0.25 +mask_time_length: + desc: null + value: 10 +mask_time_min_masks: + desc: null + value: 2 +mask_time_prob: + desc: null + value: 0.082 +max_grad_norm: + desc: null + value: 1.0 +max_length: + desc: null + value: 20 +max_steps: + desc: null + value: -1 +metric_for_best_model: + desc: null + value: None +min_length: + desc: null + value: 0 +model_type: + desc: null + value: wav2vec2 +mp_parameters: + desc: null + value: '' +no_cuda: + desc: null + value: false +no_repeat_ngram_size: + desc: null + value: 0 +num_adapter_layers: + desc: null + value: 3 +num_attention_heads: + desc: null + value: 16 +num_beam_groups: + desc: null + value: 1 +num_beams: + desc: null + value: 1 +num_codevector_groups: + desc: null + value: 2 +num_codevectors_per_group: + desc: null + value: 320 +num_conv_pos_embedding_groups: + desc: null + value: 16 +num_conv_pos_embeddings: + desc: null + value: 128 +num_feat_extract_layers: + desc: null + value: 7 +num_hidden_layers: + desc: null + value: 48 +num_negatives: + desc: null + value: 100 +num_return_sequences: + desc: null + value: 1 +num_train_epochs: + desc: null + value: 40.0 +optim: + desc: null + value: adamw_hf +output_attentions: + desc: null + value: false +output_dir: + desc: null + value: ./ +output_hidden_size: + desc: null + value: 1280 +output_hidden_states: + desc: null + value: false +output_scores: + desc: null + value: false +overwrite_output_dir: + desc: null + value: false +pad_token_id: + desc: null + value: 31 +past_index: + desc: null + value: -1 +per_device_eval_batch_size: + desc: null + value: 12 +per_device_train_batch_size: + desc: null + value: 12 +per_gpu_eval_batch_size: + desc: null + value: None +per_gpu_train_batch_size: + desc: null + value: None +prediction_loss_only: + desc: null + value: false +prefix: + desc: null + value: null +problem_type: + desc: null + value: null +proj_codevector_dim: + desc: null + value: 1024 +pruned_heads: + desc: null + value: {} +push_to_hub: + desc: null + value: true +push_to_hub_model_id: + desc: null + value: None +push_to_hub_organization: + desc: null + value: None +push_to_hub_token: + desc: null + value: +remove_invalid_values: + desc: null + value: false +remove_unused_columns: + desc: null + value: true +repetition_penalty: + desc: null + value: 1.0 +report_to: + desc: null + value: '[''wandb'']' +resume_from_checkpoint: + desc: null + value: None +return_dict: + desc: null + value: true +return_dict_in_generate: + desc: null + value: false +run_name: + desc: null + value: ./ +save_on_each_node: + desc: null + value: false +save_steps: + desc: null + value: 500 +save_strategy: + desc: null + value: steps +save_total_limit: + desc: null + value: 3 +seed: + desc: null + value: 42 +sep_token_id: + desc: null + value: null +sharded_ddp: + desc: null + value: '[]' +skip_memory_metrics: + desc: null + value: true +task_specific_params: + desc: null + value: null +tdnn_dilation: + desc: null + value: + - 1 + - 2 + - 3 + - 1 + - 1 +tdnn_dim: + desc: null + value: + - 512 + - 512 + - 512 + - 512 + - 1500 +tdnn_kernel: + desc: null + value: + - 5 + - 3 + - 3 + - 1 + - 1 +temperature: + desc: null + value: 1.0 +tf32: + desc: null + value: None +tie_encoder_decoder: + desc: null + value: false +tie_word_embeddings: + desc: null + value: true +tokenizer_class: + desc: null + value: null +top_k: + desc: null + value: 50 +top_p: + desc: null + value: 1.0 +torch_dtype: + desc: null + value: float32 +torchscript: + desc: null + value: false +tpu_metrics_debug: + desc: null + value: false +tpu_num_cores: + desc: null + value: None +train_batch_size: + desc: null + value: 12 +transformers_version: + desc: null + value: 4.18.0 +typical_p: + desc: null + value: 1.0 +use_bfloat16: + desc: null + value: false +use_legacy_prediction_loop: + desc: null + value: false +use_weighted_layer_sum: + desc: null + value: false +vocab_size: + desc: null + value: 34 +warmup_ratio: + desc: null + value: 0.0 +warmup_steps: + desc: null + value: 2000 +weight_decay: + desc: null + value: 0.0 +xpu_backend: + desc: null + value: None +xvector_output_dim: + desc: null + value: 512