Update trtllm toolkit path (#1260)

deepjavalibrary · Nov 3, 2023 · 7660014 · 7660014
1 parent 1d51b1f
commit 7660014
Showing 1 changed file with 2 additions and 3 deletions.
diff --git a/serving/docker/tensorrt-llm.Dockerfile b/serving/docker/tensorrt-llm.Dockerfile
@@ -13,16 +13,15 @@ ARG version=12.2.2-cudnn8-runtime-ubuntu22.04
 FROM nvidia/cuda:$version
 ARG cuda_version=cu122
 ARG python_version=3.10
-ARG TRT_LLM_VERSION=release/0.5.0
 ARG TORCH_VERSION=2.1.0
 ARG djl_version=0.24.0~SNAPSHOT
 ARG transformers_version=4.34.0
 ARG accelerate_version=0.23.0
 ARG tensorrtlibs_version=9.1.0.post12.dev4
-ARG trtllm_version=nightly
+ARG trtllm_toolkit_version=nightly
 ARG cuda_python_version=12.2.0
 ARG peft_wheel="https://publish.djl.ai/peft/peft-0.5.0alpha-py3-none-any.whl"
-ARG trtllm_toolkit_wheel="https://publish.djl.ai/tensorrt-llm/tensorrt_llm_toolkit-${trtllm_version}-py3-none-any.whl"
+ARG trtllm_toolkit_wheel="https://publish.djl.ai/tensorrt-llm/toolkit/tensorrt_llm_toolkit-${trtllm_toolkit_version}-py3-none-any.whl"
 ARG trtllm_wheel="https://djl-ai.s3.amazonaws.com/publish/tensorrt-llm/0.5.0/tensorrt_llm-0.5.0-py3-none-any.whl"
 ARG triton_toolkit_wheel="https://publish.djl.ai/tritonserver/r23.09/tritontoolkit-23.9-py310-none-any.whl"
 EXPOSE 8080