fix by CR bot's comments

Signed-off-by: Lizhi Zhou <[email protected]>

Author: reasonsolo

docker/Makefile

@@ -138,6 +138,8 @@ RUN_CMD           ?=
 CONTAINER_NAME    ?= tensorrt_llm
 WORK_DIR          ?= $(CODE_DIR)
 DOCKER_PULL       ?= 0
+BASE_DIR=`echo $PWD | sed 's/\//-/g'`
+GPU_OPTS=--gpus='"device=0,1,2,3,4,5,6"'
 
 %_run:
 ifeq ($(IS_ROOTLESS),1)
@@ -153,13 +155,12 @@ endif
     		$(GPU_OPTS) \
     		--volume $(SOURCE_DIR):$(CODE_DIR) \
     		$(EXTRA_VOLUMES) \
-    		$(if $(and $(filter 1,$(LOCAL_USER)),$(shell [ -w "$(USER_CACHE_DIR)" ] && echo 1)),--volume $(USER_CACHE_DIR):/home/$(USER_NAME)/.cache:rw) \
     		--env "CCACHE_DIR=$(CCACHE_DIR)" \
     		--env "CCACHE_BASEDIR=$(CODE_DIR)" \
     		--env "CONAN_HOME=$(CONAN_DIR)" \
     		--workdir $(WORK_DIR) \
     		--hostname $(shell hostname)-$* \
-    		--name $(CONTAINER_NAME)-$*-$(USER_NAME) \
+    		--name $(CONTAINER_NAME)-$*-$(USER_NAME)-$(BASE_DIR) \
     		--tmpfs /tmp:exec \
     		$(IMAGE_WITH_TAG)$(IMAGE_TAG_SUFFIX) $(RUN_CMD)
 

tests/integration/defs/accuracy/test_disaggregated_serving.py

@@ -727,19 +727,20 @@ def test_auto_dtype(self, overlap_scheduler):
 @skip_pre_blackwell
 @pytest.mark.timeout(3600)
 class TestQwen3_30B_A3B(LlmapiAccuracyTestHarness):
-    fp4_model = f"{llm_models_root()}/Qwen3/saved_models_Qwen3-30B-A3B_nvfp4_hf"
-    fp8_model = f"{llm_models_root()}/Qwen3/saved_models_Qwen3-30B-A3B_fp8_hf"
-
-    @pytest.mark.parametrize("ctxpp,gentp", [(2, 2)], ids=["ctxpp2gentp2"])
-    def test_mixed_ctx_gen_model(self, ctxpp, gentp):
-        ctx_model = self.fp4_model
-        gen_model = self.fp8_model
+    FP4_MODEL = f"{llm_models_root()}/Qwen3/saved_models_Qwen3-30B-A3B_nvfp4_hf"
+    FP8_MODEL = f"{llm_models_root()}/Qwen3/saved_models_Qwen3-30B-A3B_fp8_hf"
+
+    @pytest.mark.skip_less_device(4)
+    @pytest.mark.parametrize("ctx_pp,gen_tp", [(2, 2)], ids=["ctxpp2gentp2"])
+    def test_mixed_ctx_gen_model(self, ctx_pp, gen_tp):
+        ctx_model = self.FP4_MODEL
+        gen_model = self.FP8_MODEL
         return run_parallel_test("Qwen3/Qwen3-30B-A3B",
                                  ctx_model,
-                                 ctx_pp=ctxpp,
+                                 ctx_pp=ctx_pp,
                                  ctx_tp=1,
                                  gen_pp=1,
-                                 gen_tp=gentp,
+                                 gen_tp=gen_tp,
                                  test_sets=[GSM8K, MMLU],
                                  ctx_model=ctx_model,
                                  gen_model=gen_model)