Added model files

README.md

@@ -0,0 +1,142 @@
+---
+language: hi
+datasets:
+- common_voice 
+- indic tts
+- iiith 
+metrics:
+- wer
+tags:
+- audio
+- automatic-speech-recognition
+- speech
+- xlsr-fine-tuning-week
+license: apache-2.0
+model-index:
+- name: Hindi XLSR Wav2Vec2 Large 53
+results:
+- task: 
+    name: Speech Recognition
+    type: automatic-speech-recognition
+    dataset:
+      - name: Common Voice hi
+        type: common_voice
+        args: hi 
+      - name: Indic IIT (IITM)
+        type: indic
+        args: hi
+      - name: IIITH Indic Dataset
+        type: iiith
+        args: hi
+    metrics:
+       - name: Test WER
+         type: wer
+         value: 19.05
+---
+
+# Wav2Vec2-Large-XLSR-53-Hindi
+
+Fine-tuned [facebook/wav2vec2-large-xlsr-53](https://huggingface.co/facebook/wav2vec2-large-xlsr-53) on Hindi using the following datasets:
+- [Common Voice](https://huggingface.co/datasets/common_voice), 
+- [Indic TTS- IITM](https://www.iitm.ac.in/donlab/tts/index.php) and
+- [IIITH - Indic Speech Datasets](http://speech.iiit.ac.in/index.php/research-svl/69.html)
+
+The Hindi CommonVoice data is skewed towards male voices. However the other Indic datasets are well balanced. 
+
+Fine-tuned on facebook/wav2vec2-large-xlsr-53 using Hindi dataset :: 30 epochs >> 19.05% WER
+                          Resuming from checkpoints trained for another XX epochs >> XX.XX%
+
+When using this model, make sure that your speech input is sampled at 16kHz.
+
+## Usage
+
+The model can be used directly (without a language model) as follows:
+
+```python
+import torch
+import torchaudio
+from datasets import load_dataset
+from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
+test_dataset = load_dataset("common_voice", "hi", split="test")
+
+processor = Wav2Vec2Processor.from_pretrained("skylord/wav2vec2-large-xlsr-hindi") 
+model = Wav2Vec2ForCTC.from_pretrained("skylord/wav2vec2-large-xlsr-hindi") 
+
+resampler = torchaudio.transforms.Resample(48_000, 16_000)
+
+# Preprocessing the datasets.
+# We need to read the aduio files as arrays
+def speech_file_to_array_fn(batch):
+  speech_array, sampling_rate = torchaudio.load(batch["path"])
+  batch["speech"] = resampler(speech_array).squeeze().numpy()
+  return batch
+
+test_dataset = test_dataset.map(speech_file_to_array_fn)
+inputs = processor(test_dataset["speech"][:2], sampling_rate=16_000, return_tensors="pt", padding=True)
+
+with torch.no_grad():
+  logits = model(inputs.input_values, attention_mask=inputs.attention_mask).logits
+  
+predicted_ids = torch.argmax(logits, dim=-1)
+print("Prediction:", processor.batch_decode(predicted_ids))
+print("Reference:", test_dataset["sentence"][:2])
+```
+
+
+## Evaluation
+
+The model can be evaluated as follows on the Hindi test data of Common Voice. 
+
+
+```python
+import torch
+import torchaudio
+from datasets import load_dataset, load_metric
+from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
+import re
+
+test_dataset = load_dataset("common_voice", "el", split="test") 
+wer = load_metric("wer")
+
+processor = Wav2Vec2Processor.from_pretrained("skylord/wav2vec2-large-xlsr-hindi") 
+model = Wav2Vec2ForCTC.from_pretrained("skylord/wav2vec2-large-xlsr-hindi")
+model.to("cuda")
+
+chars_to_ignore_regex = '[\,\?\.\!\-\'\;\:\"\“\%\‘\”\�Utrnle\_]'
+unicode_ignore_regex = r'[dceMaWpmFui\xa0]' # Some unwanted unicode chars
+resampler = torchaudio.transforms.Resample(48_000, 16_000)
+
+# Preprocessing the datasets.
+# We need to read the aduio files as arrays
+
+def speech_file_to_array_fn(batch):
+  batch["sentence"] = re.sub(chars_to_ignore_regex, '', batch["sentence"]).sub(unicode_ignore_regex, '', batch["sentence"])
+  speech_array, sampling_rate = torchaudio.load(batch["path"])
+  batch["speech"] = resampler(speech_array).squeeze().numpy()
+  return batch
+  
+test_dataset = test_dataset.map(speech_file_to_array_fn)
+
+# Preprocessing the datasets.
+# We need to read the aduio files as arrays
+
+def evaluate(batch):
+  inputs = processor(batch["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)
+  with torch.no_grad():
+    logits = model(inputs.input_values.to("cuda"), attention_mask=inputs.attention_mask.to("cuda")).logits
+  pred_ids = torch.argmax(logits, dim=-1)
+  batch["pred_strings"] = processor.batch_decode(pred_ids)
+  return batch
+
+result = test_dataset.map(evaluate, batched=True, batch_size=8)
+print("WER: {:2f}".format(100 * wer.compute(predictions=result["pred_strings"], references=result["sentence"])))
+```
+
+**Test Result**: 19.056 % 
+
+
+## Training
+
+The Common Voice `train`, `validation`, datasets were used for training as well as   
+
+The script used for training & wandb dashboard can be found [here](https://wandb.ai/thinkevolve/huggingface/reports/Project-Hindi-XLSR-Large--Vmlldzo2MTI2MTQ)

config.json

@@ -0,0 +1,76 @@
+{
+  "_name_or_path": "facebook/wav2vec2-large-xlsr-53",
+  "activation_dropout": 0.0,
+  "apply_spec_augment": true,
+  "architectures": [
+    "Wav2Vec2ForCTC"
+  ],
+  "attention_dropout": 0.094,
+  "bos_token_id": 1,
+  "conv_bias": true,
+  "conv_dim": [
+    512,
+    512,
+    512,
+    512,
+    512,
+    512,
+    512
+  ],
+  "conv_kernel": [
+    10,
+    3,
+    3,
+    3,
+    3,
+    2,
+    2
+  ],
+  "conv_stride": [
+    5,
+    2,
+    2,
+    2,
+    2,
+    2,
+    2
+  ],
+  "ctc_loss_reduction": "mean",
+  "ctc_zero_infinity": false,
+  "do_stable_layer_norm": true,
+  "eos_token_id": 2,
+  "feat_extract_activation": "gelu",
+  "feat_extract_dropout": 0.0,
+  "feat_extract_norm": "layer",
+  "feat_proj_dropout": 0.0,
+  "final_dropout": 0.0,
+  "gradient_checkpointing": true,
+  "hidden_act": "gelu",
+  "hidden_dropout": 0.047,
+  "hidden_size": 1024,
+  "initializer_range": 0.02,
+  "intermediate_size": 4096,
+  "layer_norm_eps": 1e-05,
+  "layerdrop": 0.041,
+  "mask_channel_length": 10,
+  "mask_channel_min_space": 1,
+  "mask_channel_other": 0.0,
+  "mask_channel_prob": 0.0,
+  "mask_channel_selection": "static",
+  "mask_feature_length": 10,
+  "mask_feature_prob": 0.0,
+  "mask_time_length": 10,
+  "mask_time_min_space": 1,
+  "mask_time_other": 0.0,
+  "mask_time_prob": 0.082,
+  "mask_time_selection": "static",
+  "model_type": "wav2vec2",
+  "num_attention_heads": 16,
+  "num_conv_pos_embedding_groups": 16,
+  "num_conv_pos_embeddings": 128,
+  "num_feat_extract_layers": 7,
+  "num_hidden_layers": 24,
+  "pad_token_id": 93,
+  "transformers_version": "4.5.0.dev0",
+  "vocab_size": 94
+}

preprocessor_config.json

@@ -0,0 +1,8 @@
+{
+  "do_normalize": true,
+  "feature_size": 1,
+  "padding_side": "right",
+  "padding_value": 0.0,
+  "return_attention_mask": true,
+  "sampling_rate": 16000
+}

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:be0fd803682bd53a5e10531a8c98d1d154a4303d6d03d28a07c09b221c0e324e
+size 1262319255

special_tokens_map.json

@@ -0,0 +1 @@
+{"bos_token": "<s>", "eos_token": "</s>", "unk_token": "[UNK]", "pad_token": "[PAD]"}

tokenizer_config.json

@@ -0,0 +1 @@
+{"unk_token": "[UNK]", "bos_token": "<s>", "eos_token": "</s>", "pad_token": "[PAD]", "do_lower_case": false, "word_delimiter_token": "/"}

vocab.json

@@ -0,0 +1 @@
+{"प": 0, "ण": 1, "ड": 2, "फ": 3, "फ़": 4, "ं": 5, "?": 6, "ष": 7, "ँ": 8, "ृ": 9, "थ": 10, "ऊ": 11, "u": 12, "ग": 13, "r": 14, "द": 15, "र": 16, "ी": 17, "झ": 18, "ः": 19, "ब": 20, "ऐ": 21, ".": 22, "ख": 23, "घ": 24, "स": 25, "ौ": 26, "ई": 27, "F": 28, "उ": 29, "छ": 30, "ऋ": 31, "ि": 32, "े": 33, "!": 34, "ू": 35, "न": 37, "आ": 38, "ै": 39, "ऑ": 40, "भ": 41, ",": 42, "्": 43, "य": 44, "ञ": 45, "ढ": 46, "इ": 47, "।": 48, "क": 49, "ऱ": 50, "ङ": 51, "ॅ": 52, "ओ": 53, "ज़": 54, "ॉ": 55, "e": 56, "p": 57, "a": 58, "l": 59, "M": 60, "ध": 61, "ा": 62, "व": 63, "ट": 64, ":": 65, "\"": 66, "'": 67, "ढ़": 68, "ड़": 69, "ो": 70, " ": 71, "ल": 72, "अ": 73, "क़": 74, "त": 75, "ु": 76, "औ": 77, "m": 78, "श": 79, "च": 80, "़": 81, "ए": 82, "ह": 83, "W": 84, "ग़": 85, "ठ": 86, "ज": 87, "-": 88, "i": 89, "ख़": 90, "म": 91, "/": 36, "[UNK]": 92, "[PAD]": 93}