We read every piece of feedback, and take your input very seriously.
To see all available qualifiers, see our documentation.
num_logits_to_keep
torch.from_numpy()
bitsandbytes
return_dict_in_generate
llm_optims.md