Spaces:
Build error
Build error
Commit
·
55c6d0d
1
Parent(s):
35dce23
add giga am v2
Browse files
model.py
CHANGED
|
@@ -440,6 +440,7 @@ def _get_russian_pre_trained_model_ctc(
|
|
| 440 |
) -> sherpa_onnx.OfflineRecognizer:
|
| 441 |
assert repo_id in (
|
| 442 |
"csukuangfj/sherpa-onnx-nemo-ctc-giga-am-russian-2024-10-24",
|
|
|
|
| 443 |
), repo_id
|
| 444 |
|
| 445 |
model = _get_nn_model_filename(
|
|
@@ -467,6 +468,7 @@ def _get_russian_pre_trained_model(
|
|
| 467 |
"alphacep/vosk-model-ru",
|
| 468 |
"alphacep/vosk-model-small-ru",
|
| 469 |
"csukuangfj/sherpa-onnx-nemo-transducer-giga-am-russian-2024-10-24",
|
|
|
|
| 470 |
), repo_id
|
| 471 |
|
| 472 |
if repo_id == "alphacep/vosk-model-ru":
|
|
@@ -477,7 +479,10 @@ def _get_russian_pre_trained_model(
|
|
| 477 |
model_dir = "am"
|
| 478 |
encoder = "encoder.onnx"
|
| 479 |
model_type = "transducer"
|
| 480 |
-
elif repo_id
|
|
|
|
|
|
|
|
|
|
| 481 |
model_dir = "."
|
| 482 |
encoder = "encoder.int8.onnx"
|
| 483 |
model_type = "nemo_transducer"
|
|
@@ -500,7 +505,10 @@ def _get_russian_pre_trained_model(
|
|
| 500 |
subfolder=model_dir,
|
| 501 |
)
|
| 502 |
|
| 503 |
-
if repo_id
|
|
|
|
|
|
|
|
|
|
| 504 |
tokens = _get_token_filename(repo_id=repo_id, subfolder=".")
|
| 505 |
else:
|
| 506 |
tokens = _get_token_filename(repo_id=repo_id, subfolder="lang")
|
|
@@ -1861,6 +1869,8 @@ japanese_models = {
|
|
| 1861 |
}
|
| 1862 |
|
| 1863 |
russian_models = {
|
|
|
|
|
|
|
| 1864 |
"csukuangfj/sherpa-onnx-nemo-transducer-giga-am-russian-2024-10-24": _get_russian_pre_trained_model,
|
| 1865 |
"csukuangfj/sherpa-onnx-nemo-ctc-giga-am-russian-2024-10-24": _get_russian_pre_trained_model_ctc,
|
| 1866 |
"alphacep/vosk-model-ru": _get_russian_pre_trained_model,
|
|
|
|
| 440 |
) -> sherpa_onnx.OfflineRecognizer:
|
| 441 |
assert repo_id in (
|
| 442 |
"csukuangfj/sherpa-onnx-nemo-ctc-giga-am-russian-2024-10-24",
|
| 443 |
+
"csukuangfj/sherpa-onnx-nemo-ctc-giga-am-v2-russian-2025-04-19",
|
| 444 |
), repo_id
|
| 445 |
|
| 446 |
model = _get_nn_model_filename(
|
|
|
|
| 468 |
"alphacep/vosk-model-ru",
|
| 469 |
"alphacep/vosk-model-small-ru",
|
| 470 |
"csukuangfj/sherpa-onnx-nemo-transducer-giga-am-russian-2024-10-24",
|
| 471 |
+
"csukuangfj/sherpa-onnx-nemo-transducer-giga-am-v2-russian-2025-04-19",
|
| 472 |
), repo_id
|
| 473 |
|
| 474 |
if repo_id == "alphacep/vosk-model-ru":
|
|
|
|
| 479 |
model_dir = "am"
|
| 480 |
encoder = "encoder.onnx"
|
| 481 |
model_type = "transducer"
|
| 482 |
+
elif repo_id in (
|
| 483 |
+
"csukuangfj/sherpa-onnx-nemo-transducer-giga-am-russian-2024-10-24",
|
| 484 |
+
"csukuangfj/sherpa-onnx-nemo-transducer-giga-am-v2-russian-2025-04-19",
|
| 485 |
+
):
|
| 486 |
model_dir = "."
|
| 487 |
encoder = "encoder.int8.onnx"
|
| 488 |
model_type = "nemo_transducer"
|
|
|
|
| 505 |
subfolder=model_dir,
|
| 506 |
)
|
| 507 |
|
| 508 |
+
if repo_id in (
|
| 509 |
+
"csukuangfj/sherpa-onnx-nemo-transducer-giga-am-russian-2024-10-24",
|
| 510 |
+
"csukuangfj/sherpa-onnx-nemo-transducer-giga-am-v2-russian-2025-04-19",
|
| 511 |
+
):
|
| 512 |
tokens = _get_token_filename(repo_id=repo_id, subfolder=".")
|
| 513 |
else:
|
| 514 |
tokens = _get_token_filename(repo_id=repo_id, subfolder="lang")
|
|
|
|
| 1869 |
}
|
| 1870 |
|
| 1871 |
russian_models = {
|
| 1872 |
+
"csukuangfj/sherpa-onnx-nemo-transducer-giga-am-v2-russian-2025-04-19": _get_russian_pre_trained_model,
|
| 1873 |
+
"csukuangfj/sherpa-onnx-nemo-ctc-giga-am-v2-russian-2025-04-19": _get_russian_pre_trained_model_ctc,
|
| 1874 |
"csukuangfj/sherpa-onnx-nemo-transducer-giga-am-russian-2024-10-24": _get_russian_pre_trained_model,
|
| 1875 |
"csukuangfj/sherpa-onnx-nemo-ctc-giga-am-russian-2024-10-24": _get_russian_pre_trained_model_ctc,
|
| 1876 |
"alphacep/vosk-model-ru": _get_russian_pre_trained_model,
|