license: apache-2.0
Optimum Habana is the interface between the Hugging Face Transformers and Diffusers libraries and Habana's Gaudi processor (HPU). It provides a set of tools enabling easy and fast model loading, training and inference on single- and multi-HPU settings for different downstream tasks. Learn more about how to take advantage of the power of Habana HPUs to train and deploy Transformers and Diffusers models at hf.co/hardware/habana.
DistilBERT Base model HPU configuration
This model only contains the GaudiConfig file for running the distilbert-base-uncased model on Habana's Gaudi processors (HPU).
This model contains no model weights, only a GaudiConfig.
This enables to specify:
use_fused_adam: whether to use Habana's custom AdamW implementationuse_fused_clip_norm: whether to use Habana's fused gradient norm clipping operatoruse_torch_autocast: whether to use Torch Autocast for managing mixed precision
Usage
The model is instantiated the same way as in the Transformers library.
The only difference is that there are a few new training arguments specific to HPUs.
It is strongly recommended to train this model doing bf16 mixed-precision training for optimal performance and accuracy.
Here is a question-answering example script to fine-tune a model on SQuAD. You can run it with DistilBERT with the following command:
PT_HPU_LAZY_MODE=0 PT_ENABLE_INT64_SUPPORT=1 python run_qa.py \
--model_name_or_path distilbert-base-uncased \
--gaudi_config_name Habana/distilbert-base-uncased \
--dataset_name squad \
--do_train \
--do_eval \
--per_device_train_batch_size 64 \
--per_device_eval_batch_size 8 \
--learning_rate 0.0005 \
--num_train_epochs 3 \
--max_seq_length 384 \
--output_dir /tmp/squad/ \
--use_habana \
--use_lazy_mode false \
--torch_compile \
--torch_compile_backend hpu_backend \
--throughput_warmup_steps 5\
--bf16 \
--save_strategy no \
--doc_stride 128 \
--overwrite_output_dir \
Check the documentation out for more advanced usage and examples.