Habana/whisper

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.

Whisper model HPU configuration

This model only contains the GaudiConfig file for running the Whisper 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 implementation
• use_fused_clip_norm: whether to use Habana's fused gradient norm clipping operator
• use_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 sequence-to-sequence speech recognition example script to fine-tune a model. You can run it with Whisper with the following command:

python run_speech_recognition_seq2seq.py \
    --model_name_or_path="openai/whisper-small" \
    --dataset_name="mozilla-foundation/common_voice_11_0" \
    --dataset_config_name="hi" \
    --language="hindi" \
    --train_split_name="train+validation" \
    --eval_split_name="test" \
    --gaudi_config_name="Habana/whisper" \
    --max_steps="5000" \
    --output_dir="/tmp/whisper-small-hi" \
    --per_device_train_batch_size="48" \
    --per_device_eval_batch_size="2" \
    --logging_steps="25" \
    --learning_rate="1e-5" \
    --warmup_steps="500" \
    --evaluation_strategy="steps" \
    --eval_steps="1000" \
    --save_strategy="steps" \
    --save_steps="1000" \
    --generation_max_length="225" \
    --preprocessing_num_workers="1" \
    --length_column_name="input_length" \
    --max_duration_in_seconds="30" \
    --text_column_name="sentence" \
    --freeze_feature_encoder="False" \
    --group_by_length \
    --bf16 \
    --overwrite_output_dir \
    --do_train \
    --do_eval \
    --predict_with_generate \
    --use_habana \
    --use_hpu_graphs_for_inference \
    --label_features_max_length 128 \
    --dataloader_num_workers 8 \
    --throughput_warmup_steps 3

Check the documentation out for more advanced usage and examples.