A package for ONNX intermediate data dumping
Project description
onnxdumper
这是一个用于 ONNX 推理会话的 Python 包,通过额外创建一个 ONNX 推理会话加载修改了输出接口的 ONNX 模型,以便在导出 ONNX 推理中间层的输出。
安装
pip install onnxdumper
使用
import onnxruntime
import onnxdumper
# 将 onnxruntime.InferenceSession 替换为 onnxdumper.InferenceSession 即可
# session = onnxdumper.InferenceSession("path/to/your/model.onnx")
session = onnxdumper.InferenceSession("path/to/your/model.onnx")
# 准备输入数据
input_feed = {
"input_name": input_data
}
# 运行推理并导出中间层输出
output_names = ["output_name"]
results = session.run(output_names, input_feed, dump_path="path/to/save/outputs")
导出格式
导出格式为 npz,使用 numpy 进行加载,data 的 key 为输出节点的名称,value 为输出节点的数据。
import numpy as np
data = np.load("path/to/save/outputs.npz")
for key in data.keys():
print(key, data[key], data[key].shape, data[key].dtype)
节点输出的key查看方式
1. 使用 onnx 加载模型查看
model = onnx.load(path_or_bytes)
for node in model.graph.node:
print(node.output)
2. 使用 netron 查看
License
MIT
Project details
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
onnxdumper-0.7.tar.gz
(2.6 kB
view details)
Built Distribution
Filter files by name, interpreter, ABI, and platform.
If you're not sure about the file name format, learn more about wheel file names.
Copy a direct link to the current filters
File details
Details for the file onnxdumper-0.7.tar.gz.
File metadata
- Download URL: onnxdumper-0.7.tar.gz
- Upload date:
- Size: 2.6 kB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/5.1.1 CPython/3.10.12
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
7bd8511ec374e713a295b5eaa36e92fc3705e20d43b5540836e5ed273b4dbc91
|
|
| MD5 |
7176abb869d0a5aeec618df922f0902e
|
|
| BLAKE2b-256 |
e2058493fb9a84cff6621c95dab3eaeb159a39c8fe7134df3a214d1d0e63508d
|
File details
Details for the file onnxdumper-0.7-py3-none-any.whl.
File metadata
- Download URL: onnxdumper-0.7-py3-none-any.whl
- Upload date:
- Size: 2.9 kB
- Tags: Python 3
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/5.1.1 CPython/3.10.12
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
953ed271ebda09426703a09773bfbf5079209b34e4c92c8fa5d57ef39eafdc36
|
|
| MD5 |
12f4cee4e5d765fdbfa993b5a731783c
|
|
| BLAKE2b-256 |
d0af1ba868b60cdaccc19d85e61956dea9c1e54d0954304d9e04fac1c4ee7ffb
|