mmpose
mmpose copied to clipboard
open-mmlab
Final saved model
Hallo,
I have a general questions i have trained several models, i would like to know for each of the trained model where are final saved weights stored ? where I can see them
because in the work directory folder created during each of training of models there are only logs generated.
Please help.
Usually, the checkpoints are also stored in the work directory folder. Could you please share the config with this issue?
Yes, these .pth files are stored checkpoints. Usually best and final checkpoints should have the same size. If their sizes are different, you may need to check if they are stored in the same experiment, or different experiments using the same work directory.
Yes, these .pth files are stored checkpoints. Usually best and final checkpoints should have the same size. If their sizes are different, you may need to check if they are stored in the same experiment, or different experiments using the same work directory.
latest.pth is the final checkpoint ?? I have checked for the experiments done the size of latest.pth = 12 bytes in most of the cases. but size of best epoch = is not 12 bytes .Some models have 400 bytes, some 300 bytes . Is it problem if they have different sizes ??
The model size doesn't seem right. It's too small that I'm afraid the checkpoint has not been stored correctly. Is there any error information in the log? And could you please check the content of the .pth file (can be loaded by torch.load())?
There is no error in the log file. In the latest.pth it contains the skeleton info, weight,biases values, sigmas and train annotations paths.but still the size is 64 KB
latest.pth is a symlink, isn't it?
Then which are final saved weights for the trained models ?
latest.pth is a symlink itself. It points to the real checkpoint file.
Please use command ll to see the target.
I meant to ask that for each of the trained models where i can get the final saved weights? is it only best AP.pth in the working directory?
Awaiting for your reply.
Inside the work_dir, there should be .log, .log.json, epoch_.pth, best_AP_epoch_.pth, and latest.pth.
latest.pth is a symlink, pointing to a specific epoch_xx.pth (which is the real checkpoint file). best_AP_epoch_xx.pth is a copy of the best checkpoint file.
Inside the work_dir, there should be .log, .log.json, epoch.pth, best_AP_epoch__.pth, and latest.pth.
latest.pth is a symlink, pointing to a specific epoch_xx.pth (which is the real checkpoint file). best_AP_epoch_xx.pth is a copy of the best checkpoint file.
I want to see for each of the trained models for the experiments done where i can get the final saved weights? Where i can get these ?
Inside the work_dir, there should be .log, .log.json, epoch.pth, best_AP_epoch__.pth, and latest.pth. latest.pth is a symlink, pointing to a specific epoch_xx.pth (which is the real checkpoint file). best_AP_epoch_xx.pth is a copy of the best checkpoint file.
I want to see for each of the trained models for the experiments done where i can get the final saved weights? Where i can get these ?
I have done several training experiments . So for each of the trained models I want to know that what are the final saved weights ?? Thats my question.
I have a general questions i have trained several models, i would like to know for each of the trained model where are final saved weights stored ? where I can see them
FYR I have a general questions i have trained several models, i would like to know for each of the trained model where are final saved weights stored ? where I can see them
Awaiting for your reply.