Tesla A10 and cuBLAS and cuDNN 8.x libs versions #197
1001ruchka
started this conversation in
General
Replies: 3 comments 9 replies
-
|
What is "Version 4.7" & "Version 2.7"? |
Beta Was this translation helpful? Give feedback.
0 replies
-
|
|
Beta Was this translation helpful? Give feedback.
3 replies
-
That's probably an hallucination from custom initial prompt, I'll change it in the next version. |
Beta Was this translation helpful? Give feedback.
6 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment

Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
-
It turned out that depending on the version of cuBLAS and cuDNN 8.x libs - errors may appear or disappear on Tesla a10.
Version 4
Timing 02:01:56,430 - ... instead of text
Purfview-Whisper-Faster log-4.zip
Version 2
Timing 02:01:01,750 - Speech is normally recognized
Purfview-Whisper-Faster log-5.zip
At the same time, if you compare the quality of recognition of text not in English (I used Russian) - the recognition quality is higher by v4
Command line arguments: --sentence --compute_type=float16
I gave up using int8 with them Whisper makes huge long sentences. float32 is slower than float16
Whisper-Faster r172.3 version was used.
Beta Was this translation helpful? Give feedback.
All reactions