MODEL_ID propagation fix

[seungrokj]

Hi,

This is to fix the bug here:
File "/opt/conda/lib/python3.10/site-packages/text_generation_server/models/flash_causal_lm.py", line 1160, in warmup
f"tunableop_{MODEL_ID.replace('/', '-')}_tp{self.world_size}_rank{self.rank}.csv",

MODEL_ID is not properly propaged.

Please review this @OlivierDehaene @Narsil

Regards,
Seungrok

[ErikKaum]

Hi @seungrokj 👋

Thanks for the PR 👍 This seems correct.
Could you also share how to reproduce the original bug so that I can double check on my machine?

[seungrokj]

Hi @ErikKaum

Nice to meet you!

To reproduce it, in AMD machine (MI300 or MI250/MI210)

export model=NousResearch/Meta-Llama-3-8B
export volume=$PWD/data
docker run --device /dev/kfd --device /dev/dri --shm-size 1g -p 8080:80 -v $volume:/data ghcr.io/huggingface/text-generation-inference:latest-rocm --mo
del-id $model --num-shard 1

Then it will just generate error here:
File "/opt/conda/lib/python3.10/site-packages/text_generation_server/models/flash_causal_lm.py", line 1160, in warmup
f"tunableop_{MODEL_ID.replace('/', '-')}_tp{self.world_size}_rank{self.rank}.csv",

[ErikKaum]

Sorry for the slow reply @seungrokj
change will be merged in with this one: #2329

[ErikKaum]

I'll close this one as the it's in with #2329 👍