dlxj commited on
Commit ·
bc96a0e
1
Parent(s): 6191d3d
固定用 numpy==1.26.4
Browse files- readme.txt +7 -0
- requirements.txt +7 -1
readme.txt
CHANGED
|
@@ -248,6 +248,13 @@ python sou_eou_time.py
|
|
| 248 |
复用前面 CTC 生成的数据,只是加字段
|
| 249 |
|
| 250 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 251 |
|
| 252 |
|
| 253 |
它是不是真的流式输入,流式输出,为什么它是一下子打印全部结果,而不是边推理边输出部分结果 `examples\asr\asr_cache_aware_streaming\speech_to_text_cache_aware_streaming_infer.py`
|
|
|
|
| 248 |
复用前面 CTC 生成的数据,只是加字段
|
| 249 |
|
| 250 |
|
| 251 |
+
C:\Users\o\.conda\envs\NeMo\Lib\site-packages\nemo
|
| 252 |
+
已打过补丁, CTC 能正常训练 RNNT 不能。
|
| 253 |
+
|
| 254 |
+
TODO:
|
| 255 |
+
把 C:\Users\o\.conda\envs\NeMo\Lib\site-packages\nemo 移动本项目来,去掉依赖
|
| 256 |
+
|
| 257 |
+
|
| 258 |
|
| 259 |
|
| 260 |
它是不是真的流式输入,流式输出,为什么它是一下子打印全部结果,而不是边推理边输出部分结果 `examples\asr\asr_cache_aware_streaming\speech_to_text_cache_aware_streaming_infer.py`
|
requirements.txt
CHANGED
|
@@ -1,8 +1,14 @@
|
|
|
|
|
|
|
|
| 1 |
datasets==3.6.0
|
| 2 |
tabulate
|
| 3 |
joblib
|
| 4 |
tqdm
|
| 5 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 6 |
torch==2.11.0
|
| 7 |
nemo-toolkit[asr,cu13]
|
| 8 |
lhotse
|
|
|
|
| 1 |
+
--extra-index-url https://download.pytorch.org/whl/cu130
|
| 2 |
+
numpy==1.26.4
|
| 3 |
datasets==3.6.0
|
| 4 |
tabulate
|
| 5 |
joblib
|
| 6 |
tqdm
|
| 7 |
+
websockets>=12.0
|
| 8 |
+
pytest>=8.0.0
|
| 9 |
+
pytest-asyncio>=0.23.0
|
| 10 |
+
loguru>=0.7.0
|
| 11 |
+
modal>=1.3.0.post1
|
| 12 |
torch==2.11.0
|
| 13 |
nemo-toolkit[asr,cu13]
|
| 14 |
lhotse
|