voicescribe-gigaam-nvidia

GigaAM v3 models converted to ONNX format for onnx-asr.

Install onnx-asr

pip install onnx-asr[cpu,hub]

Load GigaAM v3 CTC model and recognize wav file

import onnx_asr
model = onnx_asr.load_model("gigaam-v3-ctc")
print(model.recognize("test.wav"))

Load GigaAM v3 RNN-T model and recognize wav file

import onnx_asr
model = onnx_asr.load_model("gigaam-v3-rnnt")
print(model.recognize("test.wav"))

Load GigaAM v3 E2E CTC model (with punctuation and text normalization) and recognize wav file

import onnx_asr
model = onnx_asr.load_model("gigaam-v3-e2e-ctc")
print(model.recognize("test.wav"))

Load GigaAM v3 E2E RNN-T model (with punctuation and text normalization) and recognize wav file

import onnx_asr
model = onnx_asr.load_model("gigaam-v3-e2e-rnnt")
print(model.recognize("test.wav"))

Code for models export

import gigaam
from pathlib import Path

onnx_dir = "gigaam-v3-onnx"
model_version = "v3_rnnt" # or "v3_ctc"

model = gigaam.load_model(model_version)
model.to_onnx(dir_path=onnx_dir)

with Path(onnx_dir, "v3_vocab.txt").open("wt") as f:
    for i, token in enumerate(["\u2581", *(chr(ord("а") + i) for i in range(32)), "<blk>"]):
        f.write(f"{token} {i}\n")