很棒的工作！我在尝试时遇到下面的问题，有点奇怪。
原始的error log如下：
[1692782200.924099] [d72ed04e9e58:5116 :f] vfs_fuse.c:281 UCX ERROR inotify_add_watch(/tmp) failed: No space left on device
['/app/sam/segment_anything_tensorrt/scripts', '/usr/lib/python310.zip', '/usr/lib/python3.10', '/usr/lib/python3.10/lib-dynload', '/usr/local/lib/python3.10/dist-packages', '/app/sam/segment-anything', '/usr/lib/python3/dist-packages', '.', '..']
os.path.exists(f) is True
[08/23/2023-09:16:43] [TRT] [I] [MemUsageChange] Init CUDA: CPU +15, GPU +0, now: CPU 116, GPU 19090 (MiB)
[08/23/2023-09:16:48] [TRT] [I] [MemUsageChange] Init builder kernel library: CPU +424, GPU +74, now: CPU 617, GPU 19164 (MiB)
workspace: 6
/app/sam/segment_anything_tensorrt/scripts/onnx2trt.py:323: DeprecationWarning: Use set_memory_pool_limit instead.
config.max_workspace_size = workspace * 1 << 30
weights/sam_default_prompt_mask.onnx
[08/23/2023-09:16:48] [TRT] [I] ----------------------------------------------------------------
[08/23/2023-09:16:48] [TRT] [I] Input filename: weights/sam_default_prompt_mask.onnx
[08/23/2023-09:16:48] [TRT] [I] ONNX IR version: 0.0.8
[08/23/2023-09:16:48] [TRT] [I] Opset version: 17
[08/23/2023-09:16:48] [TRT] [I] Producer name: pytorch
[08/23/2023-09:16:48] [TRT] [I] Producer version: 2.1.0
[08/23/2023-09:16:48] [TRT] [I] Domain:
[08/23/2023-09:16:48] [TRT] [I] Model version: 0
[08/23/2023-09:16:48] [TRT] [I] Doc string:
[08/23/2023-09:16:48] [TRT] [I] ----------------------------------------------------------------
[08/23/2023-09:16:48] [TRT] [W] onnx2trt_utils.cpp:374: Your ONNX model has been generated with INT64 weights, while TensorRT does not natively support INT64. Attempting to cast down to INT32.
[08/23/2023-09:16:48] [TRT] [E] [graph.cpp::symbolicExecute::539] Error Code 4: Internal Error (/OneHot: an IIOneHotLayer cannot be used to compute a shape tensor)
[08/23/2023-09:16:48] [TRT] [E] ModelImporter.cpp:771: While parsing node number 146 [Tile -> "/Tile_output_0"]:
[08/23/2023-09:16:48] [TRT] [E] ModelImporter.cpp:772: --- Begin node ---
[08/23/2023-09:16:48] [TRT] [E] ModelImporter.cpp:773: input: "/Unsqueeze_3_output_0"
input: "/Reshape_2_output_0"
output: "/Tile_output_0"
name: "/Tile"
op_type: "Tile"
[08/23/2023-09:16:48] [TRT] [E] ModelImporter.cpp:774: --- End node ---
[08/23/2023-09:16:48] [TRT] [E] ModelImporter.cpp:777: ERROR: ModelImporter.cpp:195 In function parseGraph:
[6] Invalid Node - /Tile
[graph.cpp::symbolicExecute::539] Error Code 4: Internal Error (/OneHot: an IIOneHotLayer cannot be used to compute a shape tensor)
Traceback (most recent call last):
File "/app/sam/segment_anything_tensorrt/scripts/onnx2trt.py", line 376, in <module>
export_engine_prompt_encoder_and_mask_decoder(args.sam_onnx_path)
File "/app/sam/segment_anything_tensorrt/scripts/onnx2trt.py", line 329, in export_engine_prompt_encoder_and_mask_decoder
raise RuntimeError(f'failed to load ONNX file: {onnx}')
RuntimeError: failed to load ONNX file: weights/sam_default_prompt_mask.onnx我是在docker环境用运行的，基于nvcr.io/nvidia/pytorch:23.07-py3镜像，其中sam_default_embedding.engine都可以正常导出，导出的sam_default_embedding.onnx 和 sam_default_prompt_mask.onnx 也经过sam_onnx_inference.py测试是可以正常预测的。但是在将sam_default_prompt_mask.onnx转为tensorRT engine时报错了。
很棒的工作！我在尝试时遇到下面的问题，有点奇怪。
原始的error log如下：
我是在docker环境用运行的，基于
nvcr.io/nvidia/pytorch:23.07-py3镜像，其中sam_default_embedding.engine都可以正常导出，导出的sam_default_embedding.onnx 和 sam_default_prompt_mask.onnx 也经过sam_onnx_inference.py测试是可以正常预测的。但是在将sam_default_prompt_mask.onnx转为tensorRT engine时报错了。