Instructions to use sanchit-gandhi/flax-wav2vec2-2-bart-large-cv8 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use sanchit-gandhi/flax-wav2vec2-2-bart-large-cv8 with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("automatic-speech-recognition", model="sanchit-gandhi/flax-wav2vec2-2-bart-large-cv8")# Load model directly from transformers import AutoTokenizer, AutoModelForMultimodalLM tokenizer = AutoTokenizer.from_pretrained("sanchit-gandhi/flax-wav2vec2-2-bart-large-cv8") model = AutoModelForMultimodalLM.from_pretrained("sanchit-gandhi/flax-wav2vec2-2-bart-large-cv8") - Notebooks
- Google Colab
- Kaggle
| command: | |
| - python3 | |
| - ${program} | |
| - --predict_with_generate | |
| - --do_lower_case | |
| - --do_train | |
| - --do_eval | |
| - --do_predict | |
| - ${args} | |
| method: random | |
| metric: | |
| goal: minimize | |
| name: eval/wer | |
| parameters: | |
| dataset_cache_dir: | |
| value: /home/sanchitgandhi/cache/huggingface/datasets | |
| dataset_config_name: | |
| value: all | |
| dataset_name: | |
| value: mozilla-foundation/common_voice_8_0 | |
| eval_split_name: | |
| value: validation | |
| generation_max_length: | |
| value: 200 | |
| generation_num_beams: | |
| value: 5 | |
| length_column_name: | |
| value: input_length | |
| length_penalty: | |
| distribution: log_uniform | |
| max: 0.69 | |
| min: -2.3 | |
| logging_steps: | |
| value: 25 | |
| max_duration_in_seconds: | |
| value: 20 | |
| max_target_length: | |
| value: 128 | |
| model_name_or_path: | |
| value: ./ | |
| output_dir: | |
| value: ./ | |
| per_device_eval_batch_size: | |
| value: 8 | |
| preprocessing_num_workers: | |
| value: 16 | |
| test_split_name: | |
| value: test | |
| text_column_name: | |
| value: text | |
| warmup_steps: | |
| value: 500 | |
| program: run_flax_generation_seq2seq.py | |
| project: flax-wav2vec2-2-bart-large-cv8 | |