From d98b81e26021fa8c6c42aa85a57f97c544abe94f Mon Sep 17 00:00:00 2001 From: Qiaolin Yu Date: Sat, 25 Oct 2025 15:29:14 -0700 Subject: [PATCH] Accelerate deepseek fp4 b200 ci (#11993) --- test/srt/test_deepseek_v3_fp4_4gpu.py | 5 +++++ 1 file changed, 5 insertions(+) diff --git a/test/srt/test_deepseek_v3_fp4_4gpu.py b/test/srt/test_deepseek_v3_fp4_4gpu.py index 657c0cf9c..11d3e6fd2 100644 --- a/test/srt/test_deepseek_v3_fp4_4gpu.py +++ b/test/srt/test_deepseek_v3_fp4_4gpu.py @@ -32,6 +32,8 @@ class TestDeepseekV3FP4(CustomTestCase): "flashinfer_trtllm", "--quantization", "modelopt_fp4", + "--kv-cache-dtype", + "fp8_e4m3", ] cls.process = popen_launch_server( cls.model, @@ -100,6 +102,9 @@ class TestDeepseekV3FP4MTP(CustomTestCase): "1", "--speculative-num-draft-tokens", "4", + "--kv-cache-dtype", + "fp8_e4m3", + "--enable-beta-spec", ] cls.process = popen_launch_server( cls.model,