Rob Cheung
Rob Cheung
Hi - yeah I can probably get to it some time this week. Stay tuned...
got around to bringing this project back up to speed with most dependencies today
yeah we are passing `--rm` in manually now and it is much better
Am I right in assuming that the prometheus metrics that vllm exports will just automatically get propagated to the ray serve metrics endpoint? (provided i use the `log_stats` when starting...
I am not on intel, fwiw -
@dcosson also getting this on OS X, using rbenv w/ ruby 2.2.4 vagrant 1.8.5
+1 same issue. is there a reason for the dynamic import check? seems like the qualified one is always better
also curious about the state of things here. @simon-mo @rlouf do you know the latest on this?
I got a little farther by specifying the actual `.pt` file: ``` (base) ray@6c663dea2a49:~/llm-awq/tinychat$ CUDA_VISIBLE_DEVICES=0 python vlm_demo_new.py --model-path VILA1.5-13b-AWQ/ --quant-path VILA1.5-13b-AWQ/llm/vila-1.5-13b-w4-g128-awq-v2.pt --image-file https://media.substrate.run/docs-fuji-red.jpg /home/ray/anaconda3/lib/python3.10/site-packages/transformers/utils/generic.py:441: UserWarning: torch.utils._pytree._register_pytree_node is deprecated. Please use...