Add model

README.md

@@ -0,0 +1,9 @@
+---
+tags:
+- automatic-speech-recognition
+library_name: generic
+---
+
+# pyctcdecode + Hugging Face model
+
+Inspired on https://github.com/kensho-technologies/pyctcdecode/blob/main/tutorials/02_pipeline_huggingface.ipynb

config.json

@@ -0,0 +1,77 @@
+{
+  "_name_or_path": "facebook/wav2vec2-base-960h",
+  "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": "sum",
+  "ctc_zero_infinity": false,
+  "diversity_loss_weight": 0.1,
+  "do_stable_layer_norm": false,
+  "eos_token_id": 2,
+  "feat_extract_activation": "gelu",
+  "feat_extract_dropout": 0.0,
+  "feat_extract_norm": "group",
+  "feat_proj_dropout": 0.1,
+  "feat_quantizer_dropout": 0.0,
+  "final_dropout": 0.1,
+  "gradient_checkpointing": false,
+  "hidden_act": "gelu",
+  "hidden_dropout": 0.1,
+  "hidden_dropout_prob": 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": "4.7.0.dev0",
+  "vocab_size": 32
+}

feature_extractor_config.json

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

model.py

@@ -0,0 +1,43 @@
+import numpy as np
+from typing import Dict
+
+from transformers import  Wav2Vec2Processor, Wav2Vec2ForCTC
+from pyctcdecode import Alphabet, BeamSearchDecoderCTC
+
+class PreTrainedModel():
+    def __init__(self, path):
+        """
+        Initialize model
+        """
+        processor = Wav2Vec2Processor.from_pretrained(path)
+        model = Wav2Vec2ForCTC.from_pretrained(path)
+        vocab_list = list(processor.tokenizer.get_vocab().keys())
+
+        # convert ctc blank character representation
+        vocab_list[0] = ""
+
+        # replace special characters
+        vocab_list[1] = "⁇"
+        vocab_list[2] = "⁇"
+        vocab_list[3] = "⁇"
+
+        # convert space character representation
+        vocab_list[4] = " "
+
+        alphabet = Alphabet.build_alphabet(vocab_list, ctc_token_idx=0)
+
+        self.decoder = BeamSearchDecoderCTC(alphabet)
+
+
+    def __call__(self, inputs)-> Dict[str, str]:
+        """
+        Args:
+            inputs (:obj:`np.array`):
+                The raw waveform of audio received. By default at 16KHz.
+        Return:
+            A :obj:`dict`:. The object return should be liked {"text": "XXX"} containing
+            the detected text from the input audio.
+        """
+        return {
+            "text": self.decoder.decode(logits)
+        }

preprocessor_config.json

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

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c34f9827b034a1b9141dbf6f652f8a60eda61cdf5771c9e05bfa99033c92cd96
+size 377667514

special_tokens_map.json

@@ -0,0 +1 @@
+{"bos_token": "<s>", "eos_token": "</s>", "unk_token": "<unk>", "pad_token": "<pad>"}

tf_model.h5

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:412742825972a6e2e877255ccd8b3416e618df15a7f1e5e4f736aa3632ce33b5
+size 377840624

tokenizer_config.json

@@ -0,0 +1 @@
+{"unk_token": "<unk>", "bos_token": "<s>", "eos_token": "</s>", "pad_token": "<pad>", "do_lower_case": false, "return_attention_mask": false, "do_normalize": true}

vocab.json

@@ -0,0 +1 @@
+{"<pad>": 0, "<s>": 1, "</s>": 2, "<unk>": 3, "|": 4, "E": 5, "T": 6, "A": 7, "O": 8, "N": 9, "I": 10, "H": 11, "S": 12, "R": 13, "D": 14, "L": 15, "U": 16, "M": 17, "W": 18, "C": 19, "F": 20, "G": 21, "Y": 22, "P": 23, "B": 24, "V": 25, "K": 26, "'": 27, "X": 28, "J": 29, "Q": 30, "Z": 31}