Yakuho

onnx转tflite后，推理速度变慢

3

我先使用onnx转成tensorflow pb，代码如下 ```python from onnx_tf.backend import prepare import onnx onnx_model = onnx.load("version-RFB-320.onnx") tf_rep = prepare(onnx_model) tf_rep.export_graph("version-RFB-320-tensorflow") ``` 然后我再将pb转为tflite ```python import tensorflow as tf saved_model_dir = 'version-RFB-320-tensorflow' # version-RFB-320-tensorflow 路径下含有pb和variables save_tf_model...

tensorflow version

19

Thanks for your excellent work.Do you have tensorflow verision about this code?

edge TPU版本

14

作者您好，我使用你的开源代码去运行quantization.py，成功生成了5个tflite版本，我选取全整型的hrnet_quant_int_only.tflite，在ubuntu上用edgetpu_compiler optimized/hrnet_quant_int_only.tflite进行编译，尝试使用TPU运行模型。但是我得到了这样的报错

作者你好，感谢你的开源！！ 我在使用您的shufflenetv2_0.75预训练权重，想做迁移学习的时候发现，我导入的权重值貌似不匹配，你用的网络结构与是哪个呢？ ```python from lib.core.model.shufflenet.simpleface import SimpleFace as SimpleFace_shufflenet import numpy as np pretrained_model = "./model/keypoints/variables/variables" model = SimpleFace_shufflenet() image = np.zeros(shape=(1, 160, 160, 3), dtype=np.float32) model(image) model.load_weights(pretrained_model) ``` ```...

I found that the service would report errors every time(maybe half an hour or an hour) it was running.Anyone can help me? ``` FATAL ERROR: Ineffective mark-compacts near heap limit...

It doesn't work！

3

```python from steam.client import SteamClient from csgo.client import CSGOClient client = SteamClient() cs = CSGOClient(client) client.cli_login(username="***", password="***") print("[Steam] Logged on as: %s" % client.user.name) print("[Steam] Community profile: %s" % client.steam_id.community_url)...

Python 版本

3

感謝作者的程式碼開源~~

作者你好，你的模型不错，感谢你的开源精神。我尝试使用你的模型去转为tflite的时候遇到的麻烦，转换器貌似不支持一些ops,报错如下 ``` 2021-04-08 15:37:58.650532: I tensorflow/compiler/jit/xla_gpu_device.cc:99] Not creating XLA devices, tf_xla_enable_xla_devices not set loc(callsite(callsite("map/TensorArrayV2_2@__inference_inference_5346" at "StatefulPartitionedCall@__inference_signature_wrapper_5514") at "StatefulPartitionedCall")): error: requires element_shape to be 1D tensor during TF Lite transformation pass...