ymcki
ymcki
Dear Mike, thanks for your follow up. Can you further clarify what you mean by "due to other features of this GPU"? Is this GPU missing some hardware feature that...
Thanks for your reply. Based on my research, L1 shared memory seems to be defined by compute capability version. Here is a list of compute capabilities and their L1 shared...
On paper, the GA100 A30 GPU's performance is about half the performance of A100. I remember someone said without the INT8 support, performance will be halved. Since A30 also has...
For that I mean the forward and reverse strand of the same read are concatenated together in one read such that you are seeing first half of the read and...
I am loading gemma 2 gguf model and tried to call with_structured_output NotImplementedError(). What can I do? Would install langchain-google-vertexai help?
Great! Looking forward to that in the next release.
I run this when I am in ComfyUI/custom_nodes git clone --recursive https://github.com/receyuki/comfyui-prompt-reader-node.git and this solved my problem.
Same for me. But it gets better when training get longer for Llama-3.1-8B. Not sure if we create a reward function that penalize garbage outputs can further improve this situation.
> > It's most likely an unfortunate nature of sampling - one of the goals was to enable `min_p = 0.1` and `temperature = 1.5` to stop gibberish outputs >...
My model reached 0.8/2.0 for correctness and 1.5/2.0 for format. I uploaded it to HF. I got a bit better reasoning but was offset by poorer instruction following. Maybe this...