text-generation-inference:latest-trtllm is missing dependencies to run models

[selalipop]

System Info

ghcr.io/huggingface/text-generation-inference:latest-trtllm

--executor-worker executorWorker  --model-id /workspace/mistral-engine --port 8000 --hostname 0.0.0.0 --tokenizer-name <model_id>

Returns

2024-12-18T09:34:52.744240030Z None of PyTorch, TensorFlow >= 2.0, or Flax have been found. Models won't be available and only tokenizers, configuration and file/data utilities can be used.

Information

• Docker
• The CLI directly

Tasks

• An officially supported command
• My own modifications

Reproduction

Setup ghcr.io/huggingface/text-generation-inference:latest-trtllm
It fails.

Expected behavior

It runs.