diff --git a/autotest/configs/Qwen/Qwen3.5-397B-A17B-FP8.yml b/autotest/configs/Qwen/Qwen3.5-397B-A17B-FP8.yml index cf6c32a835..c6fdda3afd 100644 --- a/autotest/configs/Qwen/Qwen3.5-397B-A17B-FP8.yml +++ b/autotest/configs/Qwen/Qwen3.5-397B-A17B-FP8.yml @@ -5,12 +5,16 @@ h: - chat - vl engine_config: + tp: 2 dp: 4 ep: 8 extra: max-batch-size: 256 cache-max-entry-count: 0.9 backends: + - name: turbomind + communicators: + - nccl - name: pytorch communicators: - nccl @@ -32,6 +36,7 @@ h: - chat - vl engine_config: + tp: 2 dp: 4 ep: 8 extra: @@ -56,6 +61,7 @@ h: - chat - vl engine_config: + tp: 2 dp: 4 ep: 8 extra: @@ -81,3 +87,30 @@ h: top-p: 0.95 chat-template-kwargs: enable_thinking: true +- model_type: + - chat + - vl + engine_config: + tp: 8 + cp: 2 + extra: + max-batch-size: 256 + cache-max-entry-count: 0.9 + backends: + - name: turbomind + communicators: + - nccl + test_coverage: + - benchmark + - evaluate + - func + - longtext_benchmark + - longtext_evaluate + - mllm_evaluate + gen_config: + temperature: 1.0 + top-k: 20 + repetition-penalty: 1.0 + top-p: 0.95 + chat-template-kwargs: + enable_thinking: true diff --git a/autotest/configs/Qwen/Qwen3.5-397B-A17B.yml b/autotest/configs/Qwen/Qwen3.5-397B-A17B.yml index 9f7a1b7ed1..0854ffcf82 100644 --- a/autotest/configs/Qwen/Qwen3.5-397B-A17B.yml +++ b/autotest/configs/Qwen/Qwen3.5-397B-A17B.yml @@ -5,12 +5,16 @@ h: - chat - vl engine_config: + tp: 2 dp: 4 ep: 8 extra: max-batch-size: 256 cache-max-entry-count: 0.7 backends: + - name: turbomind + communicators: + - nccl - name: pytorch communicators: - nccl @@ -35,6 +39,7 @@ h: - chat - vl engine_config: + tp: 2 dp: 4 ep: 8 extra: @@ -65,6 +70,7 @@ h: - chat - vl engine_config: + tp: 2 dp: 4 ep: 8 extra: @@ -88,6 +94,33 @@ h: top-p: 0.95 chat-template-kwargs: enable_thinking: true +- model_type: + - chat + - vl + engine_config: + tp: 8 + cp: 2 + extra: + max-batch-size: 256 + cache-max-entry-count: 0.7 + backends: + - name: turbomind + communicators: + - nccl + test_coverage: + - benchmark + - evaluate + - func + - longtext_benchmark + - longtext_evaluate + - mllm_evaluate + gen_config: + temperature: 1.0 + top-k: 20 + repetition-penalty: 1.0 + top-p: 0.95 + chat-template-kwargs: + enable_thinking: true ascend: - model_type: - chat