Comments (10)
Thank you for replying.
I use Python 3.7.3, which may cause this problem.
I am going to use 3.6 and let you know what happens.
from slowfast.
I use default.py and never changed.
_C.TRAIN.CHECKPOINT_TYPE = "pytorch"
Once I switch pytorch to caffe2, I succeeded in loading the pre-trained model.
Thank you so much!!
from slowfast.
Thanks for using the code.
Could you share me what is the version of python you were using to serialize the model? For PySlowFast the recommend version of python is 3.6.
Could you also clarify what do you mean by “you tried to decode in a different way”?
In PySlowFast we support loading Caffe2 or Pytorch model, and saving model In PyTorch format.
The problem I seen here seems to be you were using some different way to encode your pretrain model (different Python version, or different serialization format). If you use your own way to serialize your model, you need to change the format to loading accordingly.
from slowfast.
I just updated my comments with more details.
When loading caffe2 models, have you set CHECKPOINT_TYPE to caffe2?
I just checked the line of your error message, it seems you have modified the checkpoint.py? Could you show me what are the changes?
from slowfast.
Although I switched 3.7.3 to 3.6.9 by using a virtual environment, I still have the same problem.
I use torch torch==1.3.0 and torchvision==0.4.1.
Would you have another idea?
from slowfast.
Have you set CHECKPOINT_TYPE to caffe2?
from slowfast.
No, I haven't.
Let me try in that way.
from slowfast.
Could you show me the path to the config you were using? It seems to me that you were loading a caffe2 pretrain model as pytorch pretrain model.
from slowfast.
You might able to find more instruction regarding to loading pretrain model in GETTING_STARTED.md
from slowfast.
Thanks. Again you could find the detailed instruction of how to load model (pytorch or caffe2) in GETTING_STARTED.md
Beside that, you probably should use the path to the config I provided in the model zoo, next to the pretrain model column, where the CHECKPOINT_TYPE has already set to caffe2.
from slowfast.
Related Issues (20)
- Working Dockerfile if anyone needs
- RuntimeError: invalid unordered_map<K, T> key HOT 1
- Why is model state dictionary different in pkl file vs pyth from pytorchvideo? HOT 1
- Is it necessary to retrain the MVIT model when modifying the PATCH_KERNEL parameter during inference? HOT 1
- Output reproducibility issue while converting Mvitv2 torch model to ONNX format HOT 1
- No action detections in output video after running run_net.py for demo HOT 1
- Socket errors received while trying to run multi-GPU training HOT 1
- The error "Non-existent config key" occurs after running run_train.py. HOT 1
- OSError: Video 0 cannot be opened HOT 2
- AttributeError: module 'PIL.Image' has no attribute 'LINEAR' HOT 1
- TypeError: object of type ‘NoneType‘ has no len()
- RuntimeError: CUDA error: out of memory HOT 1
- TypeError: 'int' object is not subscriptable HOT 1
- install
- 更改数据集
- get_best_checkpoint and get_last_checkpoint?
- when train my own dataset like AVA,the accurrcy is 0
- How to turn AVA dataset format into BIT dataset format?
- AttributeError: module 'torch' has no attribute 'compiler' HOT 2
- Why did I test the slowfast model officially provided and get a very low or wrong mAp (0.039) on the ava dataset? Do you have any ideas or suggestions? HOT 1
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from slowfast.