Blake
Blake
A PR that does this correctly with the toml would be accepted.
#83 should have fixed this. Please reopen if not.
Definitely possible but it would require the code to converted to a client server model rather than what it is now. The client would be recording and the server would...
This still a problem?
> Just now installed it on Macos. installation worked properly. However, it is not printing anything - it should be real time printing right? and how to stop it ?...
Perhaps some analysis could be done on the wav audio to see if its silence or not
So this is due to hallucinations. There doesn't seem to be a way to fully fix it due to the way that the model was trained. It hallucinates when there...
The solution I developed has resolved the issue for me. Please reopen if this is still an issue.
The memory requirement increases exponentially as the sequence length gets longer. The usage of flash attention can help with this. My repo here uses Qlora with flash attention for llama...
[rank0]: Traceback (most recent call last): [rank0]: File "trl_finetune.py", line 401, in [rank0]: trainer.train(resume_from_checkpoint=args.resume_from_checkpoint) [rank0]: File "/usr/local/lib/python3.8/dist-packages/trl/trainer/sft_trainer.py", line 361, in train [rank0]: output = super().train(*args, **kwargs) [rank0]: File "/usr/local/lib/python3.8/dist-packages/transformers/trainer.py", line...