update torch version

requirements-rocm.txt

@@ -11,7 +11,7 @@ pytest-asyncio
 tensorizer>=2.9.0
 setuptools_scm>=8
 
-torch == 2.4.0
+torch == 2.4.1
 triton == 3.0.0
 flash_attn == 2.6.1
 lmslim == 0.2.0  
\ No newline at end of file

tests/entrypoints/openai/test_oot_registration.py

@@ -10,7 +10,7 @@ def run_and_test_dummy_opt_api_server(model, tp=1):
         "--gpu-memory-utilization",
         "0.10",
         "--dtype",
-        "float32",
+        "float16",# "float32",
         "--chat-template",
         str(chatml_jinja_path),
         "--load-format",
@@ -33,7 +33,7 @@ def run_and_test_dummy_opt_api_server(model, tp=1):
         )
         generated_text = completion.choices[0].message.content
         assert generated_text is not None
-        # make sure only the first token is generated
+        # make sure only the first token is generatedvim 
         rest = generated_text.replace("<s>", "")
         assert rest == ""