commit from $USER

README.md

@@ -0,0 +1,90 @@
+---
+language: vi
+datasets:
+- vlsp
+metrics:
+- wer
+tags:
+- audio
+- automatic-speech-recognition
+- speech
+license: apache-2.0
+model-index:
+- name: Wav2vec2 Base Vietnamese
+  results:
+  - task: 
+      name: Speech Recognition
+      type: automatic-speech-recognition
+    dataset:
+      name: Common Voice vi
+      type: common_voice
+      args: vi
+    metrics:
+       - name: Test WER
+         type: wer
+         value: 31.76
+---
+# Wav2Vec2-Large-XLSR-53-Vietnamese
+Fine-tuned [dragonSwing/wav2vec2-base-pretrain-vietnamese](https://huggingface.co/dragonSwing/wav2vec2-base-pretrain-vietnamese) on Vietnamese Speech Recognition task using 100h labelled data from [VSLP dataset](https://drive.google.com/file/d/1vUSxdORDxk-ePUt-bUVDahpoXiqKchMx/view?usp=sharing).
+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", "vi", split="test")
+processor = Wav2Vec2Processor.from_pretrained("dragonSwing/wav2vec2-base-vietnamese")
+model = Wav2Vec2ForCTC.from_pretrained("dragonSwing/wav2vec2-base-vietnamese")
+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 Vietnamese 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", "vi", split="test")
+wer = load_metric("wer")
+processor = Wav2Vec2Processor.from_pretrained("dragonSwing/wav2vec2-base-vietnamese")
+model = Wav2Vec2ForCTC.from_pretrained("dragonSwing/wav2vec2-base-vietnamese")
+model.to("cuda")
+chars_to_ignore_regex = r'[,?.!\-;:"“%\'�]'
+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"]).lower()
+  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=1)
+print("WER: {:2f}".format(100 * wer.compute(predictions=result["pred_strings"], references=result["sentence"])))
+```
+**Test Result**: 31.76%

config.json

@@ -0,0 +1,75 @@
+{
+  "_name_or_path": "/content/drive/MyDrive/wav2vec2-base-vn/fosd/checkpoint-96000",
+  "activation_dropout": 0.1,
+  "apply_spec_augment": true,
+  "architectures": [
+    "Wav2Vec2ForCTC"
+  ],
+  "attention_dropout": 0.1,
+  "bos_token_id": 1,
+  "codevector_dim": 256,
+  "contrastive_logits_temperature": 0.1,
+  "conv_bias": false,
+  "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": true,
+  "diversity_loss_weight": 0.1,
+  "do_stable_layer_norm": false,
+  "eos_token_id": 2,
+  "feat_extract_activation": "gelu",
+  "feat_extract_norm": "group",
+  "feat_proj_dropout": 0.0,
+  "feat_quantizer_dropout": 0.0,
+  "final_dropout": 0.1,
+  "gradient_checkpointing": true,
+  "hidden_act": "gelu",
+  "hidden_dropout": 0.1,
+  "hidden_size": 768,
+  "initializer_range": 0.02,
+  "intermediate_size": 3072,
+  "layer_norm_eps": 1e-05,
+  "layerdrop": 0.1,
+  "mask_feature_length": 10,
+  "mask_feature_prob": 0.0,
+  "mask_time_length": 10,
+  "mask_time_prob": 0.05,
+  "model_type": "wav2vec2",
+  "num_attention_heads": 12,
+  "num_codevector_groups": 2,
+  "num_codevectors_per_group": 320,
+  "num_conv_pos_embedding_groups": 16,
+  "num_conv_pos_embeddings": 128,
+  "num_feat_extract_layers": 7,
+  "num_hidden_layers": 12,
+  "num_negatives": 100,
+  "pad_token_id": 0,
+  "proj_codevector_dim": 256,
+  "transformers_version": "4.6.1",
+  "vocab_size": 98
+}

preprocessor_config.json

@@ -0,0 +1,9 @@
+{
+    "do_normalize": true,
+    "feature_extractor_type": "Wav2Vec2FeatureExtractor",
+    "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 sha256:9ec143e2442a06b362622f1c6238b3af164858f4e7dc4f4c1cda998d402c7e86
+size 377873047

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 @@
+{"<pad>": 0, "<s>": 1, "</s>": 2, "<unk>": 3, "|": 4, "j": 5, "ũ": 6, "f": 7, "v": 8, "ỗ": 9, "ạ": 10, "ể": 11, "é": 12, "è": 13, "ọ": 14, "s": 15, "ẻ": 16, "b": 17, "ữ": 18, "w": 19, "g": 20, "ì": 21, "k": 22, "ứ": 23, "ố": 24, "ở": 25, "e": 26, "à": 27, "â": 28, "á": 29, "ẵ": 30, "í": 31, "ử": 32, "ớ": 33, "ằ": 34, "ẩ": 35, "ẽ": 36, "ủ": 37, "ả": 38, "ệ": 39, "i": 40, "ă": 41, "ặ": 42, "d": 43, "ờ": 44, "ề": 45, "ồ": 46, "ừ": 47, "ổ": 48, "o": 49, "h": 50, "ấ": 51, "ẳ": 52, "ỳ": 53, "n": 54, "ụ": 55, "y": 56, "r": 57, "đ": 58, "ẫ": 59, "ỏ": 60, "ẹ": 61, "ễ": 62, "ĩ": 63, "ế": 64, "ỹ": 65, "p": 66, "ị": 67, "ộ": 68, "ã": 69, "ý": 70, "ắ": 71, "z": 72, "ô": 73, "ù": 74, "m": 75, "õ": 76, "c": 77, "t": 78, "ự": 79, "ợ": 80, "u": 81, "ê": 82, "ậ": 83, "ỡ": 84, "ỵ": 85, "ư": 86, "x": 87, "a": 88, "ó": 89, "ỉ": 90, "ỷ": 91, "l": 92, "ầ": 93, "q": 94, "ú": 95, "ò": 96, "ơ": 97}