Update README.md

README.md

@@ -1,18 +1,37 @@
 ---
+language:
+- pa
 license: apache-2.0
 tags:
 - generated_from_trainer
+- robust-speech-event
 datasets:
-- common_voice
+- mozilla-foundation/common_voice_7_0
+metrics:
+- wer
 model-index:
-- name: wav2vec2-large-xls-r-300m-pa-in
-  results: []
+- name: XLS-R-300M - Punjabi
+  results:
+  - task: 
+      type: automatic-speech-recognition
+      name: Speech Recognition
+    dataset:
+      type: mozilla-foundation/common_voice_7_0
+      name: Common Voice 7
+      args: pa-IN
+    metrics:
+      - type: wer    # Required. Example: wer
+        value: 45.611  # Required. Example: 20.90
+        name: Test WER    # Optional. Example: Test WER
+      - name: Test CER
+        type: cer
+        value: 15.584
 ---
 
 <!-- This model card has been generated automatically according to the information the Trainer had access to. You
 should probably proofread and complete it, then remove this comment. -->
 
-# wav2vec2-large-xls-r-300m-pa-in
+# XLS-R-300M - Punjabi
 
 This model is a fine-tuned version of [facebook/wav2vec2-xls-r-300m](https://huggingface.co/facebook/wav2vec2-xls-r-300m) on the common_voice dataset.
 It achieves the following results on the evaluation set:
@@ -67,3 +86,38 @@ The following hyperparameters were used during training:
 - Pytorch 1.10.0+cu111
 - Datasets 1.18.0
 - Tokenizers 0.10.3
+
+
+#### Evaluation Commands
+1. To evaluate on `mozilla-foundation/common_voice_7_0` with split `test`
+
+```bash
+python eval.py --model_id anuragshas/wav2vec2-large-xls-r-300m-pa-in --dataset mozilla-foundation/common_voice_7_0 --config pa-IN --split test
+```
+
+
+### Inference With LM
+
+```python
+import torch
+from datasets import load_dataset
+from transformers import AutoModelForCTC, AutoProcessor
+import torchaudio.functional as F
+model_id = "anuragshas/wav2vec2-large-xls-r-300m-pa-in"
+sample_iter = iter(load_dataset("mozilla-foundation/common_voice_7_0", "pa-IN", split="test", streaming=True, use_auth_token=True))
+sample = next(sample_iter)
+resampled_audio = F.resample(torch.tensor(sample["audio"]["array"]), 48_000, 16_000).numpy()
+model = AutoModelForCTC.from_pretrained(model_id)
+processor = AutoProcessor.from_pretrained(model_id)
+input_values = processor(resampled_audio, return_tensors="pt").input_values
+with torch.no_grad():
+    logits = model(input_values).logits
+transcription = processor.batch_decode(logits.numpy()).text
+# => "ਉਨ੍ਹਾਂ ਨੇ ਸਾਰੇ ਤੇਅਰਵੇ ਵੱਖਰੀ ਕਿਸਮ ਦੇ ਕੀਤੇ ਹਨ"
+```
+
+### Eval results on Common Voice 7 "test" (WER):
+
+| Without LM | With LM (run `./eval.py`) |
+|---|---|
+| 51.968 | 45.611 |