SUMIN/HuggingFace_transformer
1
0
Fork 0
Code Issues Pull Requests Actions 7 Packages Projects Releases Wiki Activity

Updates the default branch from master to main (#16326)

Browse Source 
* Updates the default branch from master to main

* Links from `master` to `main`

* Typo

* Update examples/flax/README.md

Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>

Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>
This commit is contained in:
Lysandre Debut
2022-03-23 08:46:59 +01:00
committed by GitHub
parent 7732148124
commit eca77f4719
101 changed files with 401 additions and 402 deletions
Download Patch File Download Diff File Expand all files Collapse all files

22
examples/pytorch/speech-recognition/README.md
View File

@@ -34,10 +34,10 @@ limitations under the License.
## Connectionist Temporal Classification
The script [`run_speech_recognition_ctc.py`](https://github.com/huggingface/transformers/blob/master/examples/pytorch/speech-recognition/run_speech_recognition_ctc.py) can be used to fine-tune any pretrained [Connectionist Temporal Classification Model](https://huggingface.co/docs/transformers/master/en/model_doc/auto#transformers.AutoModelForCTC) for automatic speech
The script [`run_speech_recognition_ctc.py`](https://github.com/huggingface/transformers/blob/main/examples/pytorch/speech-recognition/run_speech_recognition_ctc.py) can be used to fine-tune any pretrained [Connectionist Temporal Classification Model](https://huggingface.co/docs/transformers/main/en/model_doc/auto#transformers.AutoModelForCTC) for automatic speech
recognition on one of the [official speech recognition datasets](https://huggingface.co/datasets?task_ids=task_ids:automatic-speech-recognition) or a custom dataset.
Speech recognition models that have been pretrained in unsupervised fashion on audio data alone, *e.g.* [Wav2Vec2](https://huggingface.co/transformers/master/model_doc/wav2vec2.html), [HuBERT](https://huggingface.co/transformers/master/model_doc/hubert.html), [XLSR-Wav2Vec2](https://huggingface.co/transformers/master/model_doc/xlsr_wav2vec2.html), have shown to require only
Speech recognition models that have been pretrained in unsupervised fashion on audio data alone, *e.g.* [Wav2Vec2](https://huggingface.co/transformers/main/model_doc/wav2vec2.html), [HuBERT](https://huggingface.co/transformers/main/model_doc/hubert.html), [XLSR-Wav2Vec2](https://huggingface.co/transformers/main/model_doc/xlsr_wav2vec2.html), have shown to require only
very little annotated data to yield good performance on automatic speech recognition datasets.
In the script [`run_speech_recognition_ctc`], we first create a vocabulary from all unique characters of both the training data and evaluation data. Then, we preprocesses the speech recognition dataset, which includes correct resampling, normalization and padding. Finally, the pretrained speech recognition model is fine-tuned on the annotated speech recognition datasets using CTC loss.
@@ -58,7 +58,7 @@ If the environment variable is not set, the training script might freeze, *i.e.*
### Single GPU CTC
The following command shows how to fine-tune [XLSR-Wav2Vec2](https://huggingface.co/transformers/master/model_doc/xlsr_wav2vec2.html) on [Common Voice](https://huggingface.co/datasets/common_voice) using a single GPU in half-precision.
The following command shows how to fine-tune [XLSR-Wav2Vec2](https://huggingface.co/transformers/main/model_doc/xlsr_wav2vec2.html) on [Common Voice](https://huggingface.co/datasets/common_voice) using a single GPU in half-precision.
```bash
python run_speech_recognition_ctc.py \
@@ -93,7 +93,7 @@ of **0.35**.
### Multi GPU CTC
The following command shows how to fine-tune [XLSR-Wav2Vec2](https://huggingface.co/transformers/master/model_doc/xlsr_wav2vec2.html) on [Common Voice](https://huggingface.co/datasets/common_voice) using 8 GPUs in half-precision.
The following command shows how to fine-tune [XLSR-Wav2Vec2](https://huggingface.co/transformers/main/model_doc/xlsr_wav2vec2.html) on [Common Voice](https://huggingface.co/datasets/common_voice) using 8 GPUs in half-precision.
```bash
python -m torch.distributed.launch \
@@ -131,7 +131,7 @@ of **0.36**.
### Multi GPU CTC with Dataset Streaming
The following command shows how to use [Dataset Streaming mode](https://huggingface.co/docs/datasets/dataset_streaming.html)
to fine-tune [XLS-R](https://huggingface.co/transformers/master/model_doc/xls_r.html)
to fine-tune [XLS-R](https://huggingface.co/transformers/main/model_doc/xls_r.html)
on [Common Voice](https://huggingface.co/datasets/common_voice) using 4 GPUs in half-precision.
Streaming mode imposes several constraints on training:
@@ -245,11 +245,11 @@ they can serve as a baseline to improve upon.
## Sequence to Sequence
The script [`run_speech_recognition_seq2seq.py`](https://github.com/huggingface/transformers/blob/master/examples/pytorch/speech-recognition/run_speech_recognition_seq2seq.py) can be used to fine-tune any [Speech Sequence-to-Sequence Model](https://huggingface.co/docs/transformers/master/en/model_doc/auto#transformers.AutoModelForSpeechSeq2Seq) for automatic speech
The script [`run_speech_recognition_seq2seq.py`](https://github.com/huggingface/transformers/blob/main/examples/pytorch/speech-recognition/run_speech_recognition_seq2seq.py) can be used to fine-tune any [Speech Sequence-to-Sequence Model](https://huggingface.co/docs/transformers/main/en/model_doc/auto#transformers.AutoModelForSpeechSeq2Seq) for automatic speech
recognition on one of the [official speech recognition datasets](https://huggingface.co/datasets?task_ids=task_ids:automatic-speech-recognition) or a custom dataset.
A very common use case is to leverage a pretrained speech [encoding model](https://huggingface.co/docs/transformers/master/en/model_doc/auto#transformers.AutoModel),
*e.g.* [Wav2Vec2](https://huggingface.co/transformers/master/model_doc/wav2vec2.html), [HuBERT](https://huggingface.co/transformers/master/model_doc/hubert.html), [XLSR-Wav2Vec2](https://huggingface.co/transformers/master/model_doc/xlsr_wav2vec2.html) with a pretrained [text decoding model](https://huggingface.co/docs/transformers/master/en/model_doc/auto#transformers.AutoModel), *e.g.* [Bart](https://huggingface.co/docs/transformers/master/en/model_doc/bart#transformers.BartForCausalLM) to create a [SpeechEnocderDecoderModel](https://huggingface.co/docs/transformers/master/en/model_doc/speechencoderdecoder#speech-encoder-decoder-models).
A very common use case is to leverage a pretrained speech [encoding model](https://huggingface.co/docs/transformers/main/en/model_doc/auto#transformers.AutoModel),
*e.g.* [Wav2Vec2](https://huggingface.co/transformers/main/model_doc/wav2vec2.html), [HuBERT](https://huggingface.co/transformers/main/model_doc/hubert.html), [XLSR-Wav2Vec2](https://huggingface.co/transformers/main/model_doc/xlsr_wav2vec2.html) with a pretrained [text decoding model](https://huggingface.co/docs/transformers/main/en/model_doc/auto#transformers.AutoModel), *e.g.* [Bart](https://huggingface.co/docs/transformers/main/en/model_doc/bart#transformers.BartForCausalLM) to create a [SpeechEnocderDecoderModel](https://huggingface.co/docs/transformers/main/en/model_doc/speechencoderdecoder#speech-encoder-decoder-models).
Consequently, the warm-started Speech-Encoder-Decoder model can be fine-tuned in
this script.
@@ -314,7 +314,7 @@ Having warm-started the speech-encoder-decoder model `<your-user-name>/wav2vec2-
In the script [`run_speech_recognition_seq2seq`], we load the warm-started model,
the feature extractor, and the tokenizer, process a speech recognition dataset,
and then make use of the [`Seq2SeqTrainer`](https://huggingface.co/docs/transformers/master/en/main_classes/trainer#transformers.Seq2SeqTrainer).
and then make use of the [`Seq2SeqTrainer`](https://huggingface.co/docs/transformers/main/en/main_classes/trainer#transformers.Seq2SeqTrainer).
Note that it is important to also align the decoder's vocabulary with
the speech transcriptions of the dataset. *E.g.* the [`Librispeech`](https://huggingface.co/datasets/librispeech_asr) has only captilized letters in the transcriptions,
whereas BART was pretrained mostly on normalized text. Thus it is recommended to add
@@ -337,7 +337,7 @@ If the environment variable is not set, the training script might freeze, *i.e.*
### Single GPU Seq2Seq
The following command shows how to fine-tune [XLSR-Wav2Vec2](https://huggingface.co/transformers/master/model_doc/xlsr_wav2vec2.html) on [Common Voice](https://huggingface.co/datasets/common_voice) using a single GPU in half-precision.
The following command shows how to fine-tune [XLSR-Wav2Vec2](https://huggingface.co/transformers/main/model_doc/xlsr_wav2vec2.html) on [Common Voice](https://huggingface.co/datasets/common_voice) using a single GPU in half-precision.
```bash
python run_speech_recognition_seq2seq.py \
@@ -379,7 +379,7 @@ cross-entropy loss of **0.405** and word error rate of **0.0728**.
### Multi GPU Seq2Seq
The following command shows how to fine-tune [XLSR-Wav2Vec2](https://huggingface.co/transformers/master/model_doc/xlsr_wav2vec2.html) on [Common Voice](https://huggingface.co/datasets/common_voice) using 8 GPUs in half-precision.
The following command shows how to fine-tune [XLSR-Wav2Vec2](https://huggingface.co/transformers/main/model_doc/xlsr_wav2vec2.html) on [Common Voice](https://huggingface.co/datasets/common_voice) using 8 GPUs in half-precision.
```bash
python -m torch.distributed.launch \