Runtime EP Error narrowing_error when using CPUExecutionProvider

[seebrain]

I successfully converted meta-llama/Llama-3.2-11B-Vision model from torch to onnx but when I try to run it with OnnxRuntime using CPU EP I got the following error message. My onnx versions are
onnx 1.16.1
onnxruntime 1.20.0
onnxscript 0.1.0.dev20241104

2024-11-05 11:38:58.259321744 [I:onnxruntime:, inference_session.cc:583 TraceSessionOptions] Session Options { execution_mode:0 execution_order:DEFAULT enable_profiling:0 optimized_model_filepath:"" enable_mem_pattern:1 enable_mem_reuse:1 enable_cpu_mem_arena:1 profile_file_prefix:onnxruntime_profile_ session_logid: session_log_severity_level:0 session_log_verbosity_level:0 max_num_graph_transformation_steps:10 graph_optimization_level:3 intra_op_param:OrtThreadPoolParams { thread_pool_size: 0 auto_set_affinity: 0 allow_spinning: 1 dynamic_block_base_: 0 stack_size: 0 affinity_str: set_denormal_as_zero: 0 } inter_op_param:OrtThreadPoolParams { thread_pool_size: 0 auto_set_affinity: 0 allow_spinning: 1 dynamic_block_base_: 0 stack_size: 0 affinity_str: set_denormal_as_zero: 0 } use_per_session_threads:1 thread_pool_allow_spinning:1 use_deterministic_compute:0 config_options: { } }
2024-11-05 11:38:58.259406334 [I:onnxruntime:, inference_session.cc:483 operator()] Flush-to-zero and denormal-as-zero are off
2024-11-05 11:38:58.259423922 [I:onnxruntime:, inference_session.cc:491 ConstructorCommon] Creating and using per session threadpools since use_per_session_threads_ is true
2024-11-05 11:38:58.259436213 [I:onnxruntime:, inference_session.cc:509 ConstructorCommon] Dynamic block base set to 0
2024-11-05 11:39:19.692184563 [I:onnxruntime:, inference_session.cc:583 TraceSessionOptions] Session Options { execution_mode:0 execution_order:DEFAULT enable_profiling:0 optimized_model_filepath:"" enable_mem_pattern:1 enable_mem_reuse:1 enable_cpu_mem_arena:1 profile_file_prefix:onnxruntime_profile_ session_logid: session_log_severity_level:0 session_log_verbosity_level:0 max_num_graph_transformation_steps:10 graph_optimization_level:3 intra_op_param:OrtThreadPoolParams { thread_pool_size: 0 auto_set_affinity: 0 allow_spinning: 1 dynamic_block_base_: 0 stack_size: 0 affinity_str: set_denormal_as_zero: 0 } inter_op_param:OrtThreadPoolParams { thread_pool_size: 0 auto_set_affinity: 0 allow_spinning: 1 dynamic_block_base_: 0 stack_size: 0 affinity_str: set_denormal_as_zero: 0 } use_per_session_threads:1 thread_pool_allow_spinning:1 use_deterministic_compute:0 config_options: { } }
2024-11-05 11:39:19.692245285 [I:onnxruntime:, inference_session.cc:491 ConstructorCommon] Creating and using per session threadpools since use_per_session_threads_ is true
2024-11-05 11:39:19.692262042 [I:onnxruntime:, inference_session.cc:509 ConstructorCommon] Dynamic block base set to 0
*************** EP Error ***************
EP Error narrowing_error when using ['CPUExecutionProvider']
Falling back to ['CPUExecutionProvider'] and retrying.

---

Traceback (most recent call last):
File "/home/duyan/anaconda3/envs/llama/lib/python3.10/site-packages/onnxruntime/capi/onnxruntime_inference_collection.py", line 465, in init
self._create_inference_session(providers, provider_options, disabled_optimizers)
File "/home/duyan/anaconda3/envs/llama/lib/python3.10/site-packages/onnxruntime/capi/onnxruntime_inference_collection.py", line 528, in _create_inference_session
sess = C.InferenceSession(session_options, self._model_bytes, False, self._read_config_from_model)
RuntimeError: narrowing_error

The above exception was the direct cause of the following exception:

Traceback (most recent call last):
File "/data4/duyan/xeams/code/pipeline/infer_llama_vqa.py", line 26, in
mllama_session = ort.InferenceSession(mllama_onnx_model.SerializeToString(), sess_options=sess_options, providers=providers)
File "/home/duyan/anaconda3/envs/llama/lib/python3.10/site-packages/onnxruntime/capi/onnxruntime_inference_collection.py", line 478, in init
raise fallback_error from e
File "/home/duyan/anaconda3/envs/llama/lib/python3.10/site-packages/onnxruntime/capi/onnxruntime_inference_collection.py", line 473, in init
self._create_inference_session(self._fallback_providers, None)
File "/home/duyan/anaconda3/envs/llama/lib/python3.10/site-packages/onnxruntime/capi/onnxruntime_inference_collection.py", line 528, in _create_inference_session
sess = C.InferenceSession(session_options, self._model_bytes, False, self._read_config_from_model)
RuntimeError: narrowing_error