update

InternLM · Jan 15, 2025 · cadf231 · cadf231
1 parent f71908c
commit cadf231
Show file tree

Hide file tree

Showing 4 changed files with 21 additions and 14 deletions.
diff --git a/autotest/tools/pipeline/test_pipeline_chat_pytorch_llm.py b/autotest/tools/pipeline/test_pipeline_chat_pytorch_llm.py
@@ -255,14 +255,17 @@ def test_pipeline_chat_kvint8_tp4(config, common_case_config, model,
     ['internlm/internlm2_5-20b-chat', 'mistralai/Mixtral-8x7B-Instruct-v0.1'])
 def test_pipeline_chat_pytorch_pr(config, common_case_config, model):
     spawn_context = get_context('spawn')
+    case_config = {
+        k: v
+        for k, v in common_case_config.items() if k == 'memory_test'
+    }
     p = spawn_context.Process(target=run_pipeline_chat_test,
-                              args=(config, common_case_config, model,
-                                    'pytorch'))
+                              args=(config, case_config, model, 'pytorch'))
     p.start()
     p.join()
 
     # assert script
-    assert_pipeline_chat_log(config, common_case_config, model, 'pytorch')
+    assert_pipeline_chat_log(config, case_config, model, 'pytorch')
 
 
 @pytest.mark.order(6)

diff --git a/autotest/tools/pipeline/test_pipeline_chat_turbomind_llm.py b/autotest/tools/pipeline/test_pipeline_chat_turbomind_llm.py
@@ -227,12 +227,15 @@ def test_pipeline_chat_kvint8_tp4(config, common_case_config, model,
 ])
 def test_pipeline_chat_pr(config, common_case_config, model):
     spawn_context = get_context('spawn')
+    case_config = {
+        k: v
+        for k, v in common_case_config.items() if k == 'memory_test'
+    }
     p = spawn_context.Process(target=run_pipeline_chat_test,
-                              args=(config, common_case_config, model,
-                                    'turbomind'))
+                              args=(config, case_config, model, 'turbomind'))
     p.start()
     p.join()
-    assert_pipeline_chat_log(config, common_case_config, model, 'turbomind')
+    assert_pipeline_chat_log(config, case_config, model, 'turbomind')
 
 
 @pytest.mark.order(6)

diff --git a/autotest/tools/restful/test_restful_chat_hf_turbomind_llm.py b/autotest/tools/restful/test_restful_chat_hf_turbomind_llm.py
@@ -213,12 +213,14 @@ def test_restful_chat_kvint8_tp4(config, common_case_config, worker_id):
 ],
                          indirect=True)
 def test_restful_chat_pr(config, common_case_config):
-    run_all_step(
-        config, {
-            key: value
-            for key, value in common_case_config.items()
-            if key == 'memory_test'
-        })
+    case_config = {
+        k: v
+        for k, v in common_case_config.items() if k == 'memory_test'
+    }
+    run_all_step(config, {
+        key: value
+        for key, value in case_config.items() if key == 'memory_test'
+    })
 
 
 @pytest.mark.order(7)

diff --git a/autotest/utils/pipeline_chat.py b/autotest/utils/pipeline_chat.py
@@ -306,8 +306,7 @@ def run_pipeline_vl_chat_test(config,
         if not is_bf16_supported():
             backend_config.dtype = 'float16'
     else:
-        backend_config = TurbomindEngineConfig(tp=tp,
-                                               cache_max_entry_count=0.6)
+        backend_config = TurbomindEngineConfig(tp=tp)
 
     if 'llava' in model_case:
         backend_config.model_name = 'vicuna'