-
Line 180 here fails in stage 1 training.
https://github.com/fudan-generative-vision/champ/blob/02a9a24a9183727dcbb8eb432b46b3a19302bcb8/models/mutual_self_attention.py#L166-L186
I had to do
…
-
Hello, your work is remarkable!! I wanted to know which config/scripts are to be utilized for performing inference on unseen SAR data. I did go through all the project folders and files, I don't see a…
-
I was going to contribute to the ssh-config module but found that many of the things I wanted to add were discussed later. I frequently use proxyjump and proxycommand to forward ports and login automa…
-
When trying to resume a previous training we get this strange error:
```
src/tcmalloc.cc:284] Attempt to free invalid pointer 0x7f670fd8a100
Aborted
```
The only thing that helps is if we remove…
-
Hi there, could you report the FVD/IS results of different-sized Latte models on UCF-101?
If possible, the pre-trained checkpoints would be useful. Thanks!
-
I'm now training reward models using your code and I discovered that reward models are not logged during their creation.
https://github.com/tlc4418/llm_optimization/blob/main/configs/config_rm.yaml…
-
Hi, here is another issue about reproducibility of COIN dataset result.
I've also tried to reproduce your result of COIN dataset with using 8 A100 GPUs.
However, the evaluation result gives too low …
-
Why?
I think it would be helpful to have an overview for both devs and IxD/CD to see what data points are impacted on each user step in the manage training journey. I believe this way it will be easie…
-
In axolotl, there's a config parameter you can set:
`train_on_inputs: false`
It changes the way the loss is calculated when training a lora -> i.e. it ignores the loss on input tokens and only tra…
-
### Search before asking
- [X] I have searched the YOLOv8 [issues](https://github.com/ultralytics/ultralytics/issues) and [discussions](https://github.com/ultralytics/ultralytics/discussions) and fou…