config: (): colpali_engine.trainer.colmodel_training.ColModelTrainingConfig output_dir: !path ../../../models/biqwen2-last processor: (): colpali_engine.utils.transformers_wrappers.AllPurposeWrapper class_to_instanciate: !ext colpali_engine.models.BiQwen2Processor pretrained_model_name_or_path: "./models/colqwen2_base" # "./models/paligemma-3b-mix-448" # max_length: 50 model: (): colpali_engine.utils.transformers_wrappers.AllPurposeWrapper class_to_instanciate: !ext colpali_engine.models.BiQwen2 pretrained_model_name_or_path: "./models/colqwen2_base" torch_dtype: !ext torch.bfloat16 use_cache: false # device_map: "auto" # quantization_config: # (): transformers.BitsAndBytesConfig # load_in_4bit: true # bnb_4bit_quant_type: "nf4" # bnb_4bit_compute_dtype: "bfloat16" # bnb_4bit_use_double_quant: true dataset_loading_func: !ext colpali_engine.utils.dataset_transformation.load_train_set_detailed eval_dataset_loader: !import ../data/test_data.yaml # max_length: 50 run_eval: true add_suffix: true loss_func: (): colpali_engine.loss.bi_encoder_losses.BiPairwiseCELoss tr_args: (): transformers.training_args.TrainingArguments output_dir: null overwrite_output_dir: true num_train_epochs: 1 per_device_train_batch_size: 4 # 6 x 8 gpus = 48 batch size # gradient_accumulation_steps: 4 per_device_eval_batch_size: 4 eval_strategy: "steps" dataloader_num_workers: 8 # bf16: true save_steps: 500 logging_steps: 10 eval_steps: 100 warmup_steps: 500 learning_rate: 5e-5 save_total_limit: 1 # optim: "paged_adamw_8bit" peft_config: (): peft.LoraConfig r: 32 lora_alpha: 32 lora_dropout: 0.1 init_lora_weights: "gaussian" bias: "none" task_type: "FEATURE_EXTRACTION" target_modules: '(.*(model).*(down_proj|gate_proj|up_proj|k_proj|q_proj|v_proj|o_proj).*$|.*(custom_text_proj).*$)' # target_modules: '(.*(language_model).*(down_proj|gate_proj|up_proj|k_proj|q_proj|v_proj|o_proj).*$|.*(custom_text_proj).*$)'