AIMET-ONNX模型转precompiled_qnn_onnx模型报错：

[慢慢向上的蜗牛]

[2025-09-22T02:17:48.739+00:00] [INFO] Compiling model ID mqpew7oon [2025-09-22T02:17:48.821+00:00] [INFO] -=- Converting to QNN context binary embedded in ONNX -=- [2025-09-22T02:17:50.289+00:00] [INFO] -=- Converting ONNX graph: decompose GeLU layers -=- [2025-09-22T02:17:50.333+00:00] [INFO] Applying ONNX IR shape inference [2025-09-22T02:18:38.121+00:00] [INFO] Saving model [2025-09-22T02:18:38.348+00:00] [INFO] Uploading asset to https://tetrahub-qprod-userdata.s3-accelerate.amazonaws.com/models/mmd1k27om_4kA8FLvMun5JZICX.aimet.zip?uploadId=85K9Vllgx.QEbxQx9TRVBKrQxbEgO9mqXT8R1TOuaiE6RpJaVCvRgbBA6t4KQgTfbZMW5ye71gbd82gIsmcZXbYv2iKpZ1Hk_bkaCqASfUeNk2jEzY5OsLVUQ2szYsREZ3J9gkeAkB85Jltzqq5Bm8i8YjfK02lyhi.kmCI.pFM-&partNumber=1&A [2025-09-22T02:18:47.382+00:00] [INFO] Successfully uploaded asset with response status: 200 [2025-09-22T02:18:49.322+00:00] [INFO] Uploading asset to https://tetrahub-qprod-userdata.s3-accelerate.amazonaws.com/ [2025-09-22T02:18:49.618+00:00] [INFO] Successfully uploaded asset with response status: 204 [2025-09-22T02:18:49.618+00:00] [INFO] Successfully computed visualization graph for mmd1k27om [2025-09-22T02:18:52.876+00:00] [INFO] -=- ONNX to QNN DLC (qairt-converter) -=- [2025-09-22T02:18:52.876+00:00] [INFO] Running /tetra/qairt_env/default/2.37.0/bin/python3 /qairt_sdk/default/2.37.0/bin/x86_64-linux-clang/qairt-converter --input_network /tmp/tmpiga0rjqy/tmp1jdrv_l1.onnx --output_path /tmp/tmpiga0rjqy/graph_k6few90m.dlc --preserve_io_datatype logit_ids --preserve_io_datatype input_ids attention_mask position_ids logit_ids --quantization_overrides /tmp/tmp9khg85gy/qwen_aimet_mqpew7oon.aimet/qwen2.5_0.5b_aimet.encodings --onnx_skip_simplification [2025-09-22T02:19:12.205+00:00] [INFO] 2025-09-22 02:18:53,583 - 278 - INFO - INFO_INITIALIZATION_SUCCESS: | 2025-09-22 02:18:53,779 - 283 - WARNING - Unable to register converter supported Operation [Inverse:Version 1] with your Onnx installation. Converter will bail if Model contains this Op. | 2025-09-22 02:18:53,949 - 283 - WARNING - --desired_input_shape and -d are deprecated. Use --source_model_input_shape or -s for achieving this functionality | 2025-09-22 02:18:53,950 - 278 - INFO - Processing user provided quantization encodings: | 2025-09-22 02:18:53,953 - 278 - INFO - Input shape info | 2025-09-22 02:18:53,971 - 283 - WARNING - Skipping Native checker in Defer Loading and ONNX Simplification not invoked | 2025-09-22 02:18:54,186 - 283 - WARNING - Only numerical type cast is supported. The cast op: /model/rotary_emb/Cast_1 will be interpreted at conversion time | 2025-09-22 02:18:54,187 - 283 - WARNING - Only numerical type cast is supported. The cast op: /model/Cast_1 will be interpreted at conversion time | 2025-09-22 02:18:55,291 - 283 - WARNING - Only numerical type cast is supported. The cast op: /model/Cast_3 will be interpreted at conversion time | 2025-09-22 02:18:55,314 - 283 - WARNING - Only numerical type cast is supported. The cast op: /model/Cast_4 will be interpreted at conversion time | 2025-09-22 02:18:57,680 - 278 - INFO - INFO_STATIC_RESHAPE: Applying static reshape to model.embed_tokens.weight: new name model.embed_tokens.weight_transposed new shape (896, 151936) | 2025-09-22 02:18:57,916 - 283 - WARNING - Unused Input nodes found: [] | 2025-09-22 02:18:58,224 - 278 - INFO - Processed 1514 quantization encodings | 2025-09-22 02:19:11,026 - 283 - WARNING - Unused Input nodes found: [] | 2025-09-22 02:19:11,028 - 278 - INFO - INFO_INITIALIZATION_SUCCESS: | IrQuantizer: Param Quantizer should be set to symmetric for 32 bit biases. Will ignore param quantizer option: tf for biases | [ WARNING ] Following OPs fallback to float: | /model/layers.2/self_attn/Expand_1, /model/layers.2/self_attn/Reshape_4, /model/layers.2/self_attn/Expand, /model/layers.2/self_attn/Reshape_3, /model/layers.2/self_attn/Transpose_3, /model/layers.2/self_attn/Mul_7, /model/layers.2/self_attn/MatMul_1, /model/layers.0/self_attn/Reshape_3, /model/layers.0/self_attn/Expand, /model/rotary_emb/Cast_1, /model/Add, /model/layers.0/self_attn/Expand_1, /model/Mul_6, /model/layers.0/self_attn/Reshape_4, /model/Cast_1, /model/layers.0/self_attn/Mul_7, /model/layers.0/self_attn/MatMul_1, /model/rotary_emb/MatMul, /model/layers.0/self_attn/Transpose_3, /model/Cast_4, /model/layers.5/self_attn/Expand_1, /model/layers.5/self_attn/Reshape_4, /model/layers.5/self_attn/Expand, /model/layers.5/self_attn/Reshape_3, /model/layers.5/self_attn/Transpose_3, /model/layers.5/self_attn/Mul_7, /model/layers.5/self_attn/MatMul_1, /model/layers.6/self_attn/Expand_1, /model/layers.6/self_attn/Reshape_4, /model/layers.6/self_attn/Expand, /model/layers.6/self_attn/Reshape_3, /model/layers.6/self_attn/Transpose_3, /model/layers.6/self_attn/Mul_7, /model/layers.6/self_attn/MatMul_1, /model/layers.7/self_attn/Expand_1, /model/layers.7/self_attn/Reshape_4, /model/layers.7/self_attn/Expand, /model/layers.7/self_attn/Reshape_3, /model/layers.7/self_attn/Transpose_3, /model/layers.7/self_attn/Mul_7, /model/layers.7/self_attn/MatMul_1, /model/layers.8/self_attn/Expand_1, /model/layers.8/self_attn/Reshape_4, /model/layers.8/self_attn/Expand, /model/layers.8/self_attn/Reshape_3, /model/layers.8/self_attn/Transpose_3, /model/layers.8/self_attn/Mul_7, /model/layers.8/self_attn/MatMul_1, /model/layers.9/self_attn/Expand_1, /model/layers.9/self_attn/Reshape_4, /model/layers.9/self_attn/Expand, /model/layers.9/self_attn/Reshape_3, /model/layers.9/self_attn/Transpose_3, /model/layers.9/self_attn/Mul_7, /model/layers.9/self_attn/MatMul_1, /model/layers.10/self_attn/Expand_1, /model/layers.10/self_attn/Reshape_4, /model/layers.10/self_attn/Expand, /model/layers.10/self_attn/Reshape_3, /model/layers.10/self_attn/Transpose_3, /model/layers.10/self_attn/Mul_7, /model/layers.10/self_attn/MatMul_1, /model/layers.11/self_attn/Expand_1, /model/layers.11/self_attn/Reshape_4, /model/layers.11/self_attn/Expand, /model/layers.11/self_attn/Reshape_3, /model/layers.11/self_attn/Transpose_3, /model/layers.11/self_attn/Mul_7, /model/layers.11/self_attn/MatMul_1, /model/layers.3/self_attn/Expand_1, /model/layers.3/self_attn/Reshape_4, /model/layers.3/self_attn/Expand, /model/layers.3/self_attn/Reshape_3, /model/layers.3/self_attn/Transpose_3, /model/layers.3/self_attn/Mul_7, /model/layers.3/self_attn/MatMul_1, /model/layers.4/self_attn/Expand_1, /model/layers.4/self_attn/Reshape_4, /model/layers.4/self_attn/Expand, /model/layers.4/self_attn/Reshape_3, /model/layers.4/self_attn/Transpose_3, /model/layers.4/self_attn/Mul_7, /model/layers.4/self_attn/MatMul_1, /model/layers.12/self_attn/Expand_1, /model/layers.12/self_attn/Reshape_4, /model/layers.12/self_attn/Expand, /model/layers.12/self_attn/Reshape_3, /model/layers.12/self_attn/Transpose_3, /model/layers.12/self_attn/Mul_7, /model/layers.12/self_attn/MatMul_1, /model/layers.13/self_attn/Expand_1, /model/layers.13/self_attn/Reshape_4, /model/layers.13/self_attn/Expand, /model/layers.13/self_attn/Reshape_3, /model/layers.13/self_attn/Transpose_3, /model/layers.13/self_attn/Mul_7, /model/layers.13/self_attn/MatMul_1, /model/layers.14/self_attn/Expand_1, /model/layers.14/self_attn/Reshape_4, /model/layers.14/self_attn/Expand, /model/layers.14/self_attn/Reshape_3, /model/layers.14/self_attn/Transpose_3, /model/layers.14/self_attn/Mul_7, /model/layers.14/self_attn/MatMul_1, /model/layers.15/self_attn/Expand_1, /model/layers.15/self_attn/Reshape_4, /model/layers.15/self_attn/Expand, /model/layers.15/self_attn/Reshape_3, /model/layers.15/self_attn/Transpose_3, /model/layers.15/self_attn/Mul_7, /model/layers.15/self_attn/MatMul_1, /model/layers.16/self_attn/Expand_1, /model/layers.16/self_attn/Reshape_4, /model/layers.16/self_attn/Expand, /model/layers.16/self_attn/Reshape_3, /model/layers.16/self_attn/Transpose_3, /model/layers.16/self_attn/Mul_7, /model/layers.16/self_attn/MatMul_1, /model/layers.17/self_attn/Expand_1, /model/layers.17/self_attn/Reshape_4, /model/layers.17/self_attn/Expand, /model/layers.17/self_attn/Reshape_3, /model/layers.17/self_attn/Transpose_3, /model/layers.17/self_attn/Mul_7, /model/layers.17/self_attn/MatMul_1, /model/layers.18/self_attn/Expand_1, /model/layers.18/self_attn/Reshape_4, /model/layers.18/self_attn/Expand, /model/layers.18/self_attn/Reshape_3, /model/layers.18/self_attn/Transpose_3, /model/layers.18/self_attn/Mul_7, /model/layers.18/self_attn/MatMul_1, /model/layers.19/self_attn/Expand_1, /model/layers.19/self_attn/Reshape_4, /model/layers.19/self_attn/Expand, /model/layers.19/self_attn/Reshape_3, /model/layers.19/self_attn/Transpose_3, /model/layers.19/self_attn/Mul_7, /model/layers.19/self_attn/MatMul_1, /model/layers.20/self_attn/Expand_1, /model/layers.20/self_attn/Reshape_4, /model/layers.20/self_attn/Expand, /model/layers.20/self_attn/Reshape_3, /model/layers.20/self_attn/Transpose_3, /model/layers.20/self_attn/Mul_7, /model/layers.20/self_attn/MatMul_1, /model/layers.21/self_attn/Expand_1, /model/layers.21/self_attn/Reshape_4, /model/layers.21/self_attn/Expand, /model/layers.21/self_attn/Reshape_3, /model/layers.21/self_attn/Transpose_3, /model/layers.21/self_attn/Mul_7, /model/layers.21/self_attn/MatMul_1, /model/layers.1/self_attn/Expand_1, /model/layers.1/self_attn/Reshape_4, /model/layers.1/self_attn/Expand, /model/layers.1/self_attn/Reshape_3, /model/layers.1/self_attn/Transpose_3, /model/layers.1/self_attn/Mul_7, /model/layers.1/self_attn/MatMul_1, /model/layers.22/self_attn/Expand_1, /model/layers.22/self_attn/Reshape_4, /model/layers.22/self_attn/Expand, /model/layers.22/self_attn/Reshape_3, /model/layers.22/self_attn/Transpose_3, /model/layers.22/self_attn/Mul_7, /model/layers.22/self_attn/MatMul_1, /model/layers.23/self_attn/Expand_1, /model/layers.23/self_attn/Reshape_4, /model/layers.23/self_attn/Expand, /model/layers.23/self_attn/Reshape_3, /model/layers.23/self_attn/Transpose_3, /model/layers.23/self_attn/Mul_7, /model/layers.23/self_attn/MatMul_1, /lm_head/MatMul. | Fatal Python error: Segmentation fault | | Current thread 0x00007fad57600280 (most recent call first): | File "/qairt_sdk/default/2.37.0/lib/python/qti/aisw/converters/qnn_backend/ir_to_dlc.py", line 491 in serialize | File "/qairt_sdk/default/2.37.0/bin/x86_64-linux-clang/qairt-converter", line 366 in main | File "/qairt_sdk/default/2.37.0/bin/x86_64-linux-clang/qairt-converter", line 421 in <module> | | Extension modules: numpy.core._multiarray_umath, numpy.core._multiarray_tests, numpy.linalg._umath_linalg, numpy.fft._pocketfft_internal, numpy.random._common, numpy.random.bit_generator, numpy.random._bounded_integers, numpy.random._mt19937, numpy.random.mtrand, numpy.random._philox, numpy.random._pcg64, numpy.random._sfc64, numpy.random._generator, google.protobuf.pyext._message (total: 14) [2025-09-22T02:19:12.402+00:00] [INFO] QAIRT converter failed with exit code -11


有没有大佬知道如何解决这个错误