From 641c09714922d27fb02167ba396bbe90fd997d7a Mon Sep 17 00:00:00 2001 From: Suma Kasa Date: Wed, 19 Nov 2025 10:23:33 -0800 Subject: [PATCH 1/3] Install vllm 0.11.1 --- serving/docker/lmi-container-requirements.txt | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/serving/docker/lmi-container-requirements.txt b/serving/docker/lmi-container-requirements.txt index 09b41ec30..06eacda5b 100644 --- a/serving/docker/lmi-container-requirements.txt +++ b/serving/docker/lmi-container-requirements.txt @@ -32,7 +32,7 @@ uvloop ninja peft llmcompressor -https://vllm-wheels.s3.us-west-2.amazonaws.com/93103575ce0480f36fc1a3603eb51d9a89f38a00/vllm-1.0.0.dev-cp38-abi3-manylinux1_x86_64.whl +vllm@git+https://github.com/vllm-project/vllm.git@v0.11.1 xgrammar flashinfer-python==0.5.2 lmcache \ No newline at end of file From a132fd882edb3025276f6e17f999a873f7a6a87e Mon Sep 17 00:00:00 2001 From: Suma Kasa Date: Wed, 19 Nov 2025 11:03:29 -0800 Subject: [PATCH 2/3] Install vllm 0.11.1 directly from pip --- serving/docker/lmi-container-requirements.txt | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/serving/docker/lmi-container-requirements.txt b/serving/docker/lmi-container-requirements.txt index 06eacda5b..dbb2dabad 100644 --- a/serving/docker/lmi-container-requirements.txt +++ b/serving/docker/lmi-container-requirements.txt @@ -32,7 +32,7 @@ uvloop ninja peft llmcompressor -vllm@git+https://github.com/vllm-project/vllm.git@v0.11.1 +vllm==0.11.1 xgrammar flashinfer-python==0.5.2 lmcache \ No newline at end of file From 03e26b4ea7b226e22508042ddc7391d8f78b390b Mon Sep 17 00:00:00 2001 From: Suma Kasa Date: Wed, 19 Nov 2025 13:18:54 -0800 Subject: [PATCH 3/3] Fix Llama3.1 8b Speculative Eagle test --- tests/integration/llm/prepare.py | 1 + 1 file changed, 1 insertion(+) diff --git a/tests/integration/llm/prepare.py b/tests/integration/llm/prepare.py index 8a81a7000..71ae24bf7 100644 --- a/tests/integration/llm/prepare.py +++ b/tests/integration/llm/prepare.py @@ -162,6 +162,7 @@ '{"method":"eagle","model":"yuhuili/EAGLE-LLaMA3.1-Instruct-8B","num_speculative_tokens":4}', "option.tensor_parallel_degree": 4, "option.max_rolling_batch_size": 4, + "option.enforce_eager": True, }, "llama-7b-unmerged-lora": { "option.model_id": "s3://djl-llm/huggyllama-llama-7b",