search

espressif / esp-dl

Espressif deep-learning library for AIoT applications
MIT License
549 stars 118 forks source link

参照官方提供的使用 ESP-DL 深度学习库基于 ESP32-S3 实现手势识别的这篇文章中在2.3评估里使用了里面的代码但是报错（前面完全按照示例走） (AIV-693) #162

Closed emoof1 closed 5 months ago

emoof1 commented 5 months ago

错误信息 [outputs, _] = eva.evalaute_quantized_model(test_images[i batch_size:(i + 1) batch_size], False) AttributeError: 'Evaluator' object has no attribute 'evalaute_quantized_model'

Sandra-lol commented 5 months ago

请问是什么文章？

emoof1 commented 5 months ago

请问是什么文章？

image

emoof1 commented 5 months ago

请问是什么文章？

https://zhuanlan.zhihu.com/p/590495368

BlueSkyB commented 5 months ago

错误信息 [outputs, _] = eva.evalaute_quantized_model(test_images[i batch_size:(i + 1) batch_size], False) AttributeError: 'Evaluator' object has no attribute 'evalaute_quantized_model'

DL 里面这个接口名应该是'evalute_quantized_model'

emoof1 commented 5 months ago

错误信息 [outputs, _] = eva.evalaute_quantized_model(test_images[i batch_size:(i + 1) batch_size], False) AttributeError: 'Evaluator' object has no attribute 'evalaute_quantized_model'

DL 里面这个接口名应该是'evalute_quantized_model'

Traceback (most recent call last): File "H:\opencvprojiect\ESP-test\change\评估.py", line 33, in [outputs, _] = eva.evalute_quantized_model(test_images[i batch_size:(i + 1) batch_size], True) File "evaluator.py", line 1173, in evaluator.Evaluator.evalute_quantized_model File "D:\python\lib\site-packages\onnxruntime\capi\onnxruntime_inference_collection.py", line 280, in init self._create_inference_session(providers, provider_options) File "D:\python\lib\site-packages\onnxruntime\capi\onnxruntime_inference_collection.py", line 309, in _create_inference_session sess = C.InferenceSession(session_options, self._model_bytes, False, self._read_config_from_model) onnxruntime.capi.onnxruntime_pybind11_state.InvalidGraph: [ONNXRuntimeError] : 10 : INVALID_GRAPH : This is an invalid model. Error in Node:StatefulPartitionedCall/sequential/conv2d/BiasAdd/QACC_clip0 : Node (StatefulPartitionedCall/sequential/conv2d/BiasAdd/QACC_clip0) has input size 3 not in range [min=1, max=1]. 但是出现了新的错误

BlueSkyB commented 5 months ago

错误信息 [outputs, _] = eva.evalaute_quantized_model(test_images[i batch_size:(i + 1) batch_size], False) AttributeError: 'Evaluator' object has no attribute 'evalaute_quantized_model'

DL 里面这个接口名应该是'evalute_quantized_model'

Traceback (most recent call last): File "H:\opencvprojiect\ESP-test\change\评估.py", line 33, in [outputs, _] = eva.evalute_quantized_model(test_images[i batch_size:(i + 1) batch_size], True) File "evaluator.py", line 1173, in evaluator.Evaluator.evalute_quantized_model File "D:\python\lib\site-packages\onnxruntime\capi\onnxruntime_inference_collection.py", line 280, in init self._create_inference_session(providers, provider_options) File "D:\python\lib\site-packages\onnxruntime\capi\onnxruntime_inference_collection.py", line 309, in _create_inference_session sess = C.InferenceSession(session_options, self._model_bytes, False, self._read_config_from_model) onnxruntime.capi.onnxruntime_pybind11_state.InvalidGraph: [ONNXRuntimeError] : 10 : INVALID_GRAPH : This is an invalid model. Error in Node:StatefulPartitionedCall/sequential/conv2d/BiasAdd/QACC_clip0 : Node (StatefulPartitionedCall/sequential/conv2d/BiasAdd/QACC_clip0) has input size 3 not in range [min=1, max=1]. 但是出现了新的错误

这个提示是说 clip 算子，有三个输入，不符合 一个输入的要求(DL 中构建Clip算子是按三个输入构建ONNX Node的)。可以检查下你的ONNX以及ONNXRUNTIME版本，新的ONNX Clip算子定义是按三个输入来的。

emoof1 commented 5 months ago

错误信息 [outputs, _] = eva.evalaute_quantized_model(test_images[i batch_size:(i + 1) batch_size], False) AttributeError: 'Evaluator' object has no attribute 'evalaute_quantized_model'

DL 里面这个接口名应该是'evalute_quantized_model'

Traceback (most recent call last): File "H:\opencvprojiect\ESP-test\change\评估.py", line 33, in [outputs, _] = eva.evalute_quantized_model(test_images[i batch_size:(i + 1) batch_size], True) File "evaluator.py", line 1173, in evaluator.Evaluator.evalute_quantized_model File "D:\python\lib\site-packages\onnxruntime\capi\onnxruntime_inference_collection.py", line 280, in init self._create_inference_session(providers, provider_options) File "D:\python\lib\site-packages\onnxruntime\capi\onnxruntime_inference_collection.py", line 309, in _create_inference_session sess = C.InferenceSession(session_options, self._model_bytes, False, self._read_config_from_model) onnxruntime.capi.onnxruntime_pybind11_state.InvalidGraph: [ONNXRuntimeError] : 10 : INVALID_GRAPH : This is an invalid model. Error in Node:StatefulPartitionedCall/sequential/conv2d/BiasAdd/QACC_clip0 : Node (StatefulPartitionedCall/sequential/conv2d/BiasAdd/QACC_clip0) has input size 3 not in range [min=1, max=1]. 但是出现了新的错误

这个提示是说 clip 算子，有三个输入，不符合 一个输入的要求(DL 中构建Clip算子是按三个输入构建ONNX Node的)。可以检查下你的ONNX以及ONNXRUNTIME版本，新的ONNX Clip算子定义是按三个输入来的。

image 看过版本号了和官方的一样 但是int8的那个就是跑不起来也不知道怎么改

image

BlueSkyB commented 5 months ago

我重新创建新的环境，下载依赖库，使用你参考的知乎博文使用的代码和onnx，是没有问题的。(博文里有给链接：https://github.com/alibukharai/Blogs/tree/main/ESP-DL/esp_dl_formate_conversion) 所以库没有问题，你再检查下自己的环境和代码。

emoof1 commented 5 months ago

我重新创建新的环境，下载依赖库，使用你参考的知乎博文使用的代码和onnx，是没有问题的。(博文里有给链接：https://github.com/alibukharai/Blogs/tree/main/ESP-DL/esp_dl_formate_conversion) 所以库没有问题，你再检查下自己的环境和代码。

okk 感谢感谢 我再看看 我还有一个问题 就是如果要处理RGB彩图的我该怎么修改已知我的数据集是RGB24位的 但是设备里的彩图是RGB565的 有什么合理的解决方案吗