kiran

Results 6 comments of


                                            kiran

Inference on GPU

Flexgen only supports opt models

Issue with raylet error

clean up disk and keep it under 95% usage, that should fix the issue

finetuning with error

@HarrywillDr v100 gpus does not support TF32, remove that tag

OOM error while training llama-7b with five V100-32G GPUs

maybe try `full_shard offload auto_wrap`

[New Model] Support LLaMA 4 family

is there any progress on this?

[Bug]: lora response's moel name is incorrect

seeing the same issue