export-onnx-ctc-non-streaming.py
2.6 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
#!/usr/bin/env python3
# Copyright 2024 Xiaomi Corp. (authors: Fangjun Kuang)
import argparse
from typing import Dict
import nemo.collections.asr as nemo_asr
import onnx
import torch
from onnxruntime.quantization import QuantType, quantize_dynamic
def get_args():
parser = argparse.ArgumentParser()
parser.add_argument(
"--model",
type=str,
required=True,
)
parser.add_argument(
"--doc",
type=str,
default="",
)
return parser.parse_args()
def add_meta_data(filename: str, meta_data: Dict[str, str]):
"""Add meta data to an ONNX model. It is changed in-place.
Args:
filename:
Filename of the ONNX model to be changed.
meta_data:
Key-value pairs.
"""
model = onnx.load(filename)
while len(model.metadata_props):
model.metadata_props.pop()
for key, value in meta_data.items():
meta = model.metadata_props.add()
meta.key = key
meta.value = str(value)
onnx.save(model, filename)
@torch.no_grad()
def main():
args = get_args()
model_name = args.model
asr_model = nemo_asr.models.ASRModel.from_pretrained(model_name=model_name)
print(asr_model.cfg)
print(asr_model)
with open("./tokens.txt", "w", encoding="utf-8") as f:
for i, s in enumerate(asr_model.joint.vocabulary):
f.write(f"{s} {i}\n")
f.write(f"<blk> {i+1}\n")
print("Saved to tokens.txt")
decoder_type = "ctc"
asr_model.change_decoding_strategy(decoder_type=decoder_type)
asr_model.eval()
asr_model.set_export_config({"decoder_type": "ctc"})
filename = "model.onnx"
asr_model.export(filename)
normalize_type = asr_model.cfg.preprocessor.normalize
if normalize_type == "NA":
normalize_type = ""
meta_data = {
"vocab_size": asr_model.decoder.vocab_size,
"normalize_type": normalize_type,
"subsampling_factor": 8,
"model_type": "EncDecHybridRNNTCTCBPEModel",
"version": "1",
"model_author": "NeMo",
"url": f"https://catalog.ngc.nvidia.com/orgs/nvidia/teams/nemo/models/{model_name}"
if "/" in model_name
else f"https://huggingface.co/{model_name}",
"comment": "Only the CTC branch is exported",
"doc": args.doc,
}
add_meta_data(filename, meta_data)
quantize_dynamic(
model_input="./model.onnx",
model_output="./model.int8.onnx",
weight_type=QuantType.QUInt8,
)
print("preprocessor", asr_model.cfg.preprocessor)
print(meta_data)
if __name__ == "__main__":
main()