[TRTLLM] Bump up trtllm to version 0.7.1 (#1452)

deepjavalibrary · Jan 5, 2024 · cc337e1 · cc337e1
1 parent 6fcda33
commit cc337e1
Show file tree

Hide file tree

Showing 2 changed files with 3 additions and 3 deletions.
diff --git a/serving/docker/tensorrt-llm.Dockerfile b/serving/docker/tensorrt-llm.Dockerfile
@@ -19,11 +19,11 @@ ARG transformers_version=4.36.2
 ARG accelerate_version=0.25.0
 ARG tensorrtlibs_version=9.2.0.post12.dev5
 ARG trtllm_toolkit_version=nightly
-ARG trtllm_version=v0.7.0
+ARG trtllm_version=v0.7.1
 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/toolkit/tensorrt_llm_toolkit-${trtllm_toolkit_version}-py3-none-any.whl"
-ARG trtllm_wheel="https://djl-ai.s3.amazonaws.com/publish/tensorrt-llm/${trtllm_version}/tensorrt_llm-0.7.0-cp310-cp310-linux_x86_64.whl"
+ARG trtllm_wheel="https://djl-ai.s3.amazonaws.com/publish/tensorrt-llm/${trtllm_version}/tensorrt_llm-0.7.1-cp310-cp310-linux_x86_64.whl"
 ARG triton_toolkit_wheel="https://publish.djl.ai/tritonserver/r23.11/tritontoolkit-23.11-py310-none-any.whl"
 ARG pydantic_version=1.10.13
 ARG ammo_version=0.5.0

diff --git a/tests/integration/llm/prepare.py b/tests/integration/llm/prepare.py
@@ -687,7 +687,7 @@
         "option.output_formatter": "jsonlines",
     },
     "falcon-7b": {
-        "option.model_id": "s3://djl-llm/triton/0.7.0/falcon-7b-tp1-bs16/",
+        "option.model_id": "s3://djl-llm/triton/0.7.1/falcon-7b-tp1-bs16/",
         "option.tensor_parallel_degree": 1,
         "option.max_input_len": 1024,
         "option.max_output_len": 512,