-
Cloudflare R2 is cheaper than S3 (no egress fees) and is very easy to use.
It is fully S3 compatible, so it should be possible to just use the S3 object store logger (if I can specify the Cloudflare …
-
Dear Yuan, thank you for releasing the source code for your great project. It looks like the current code only supports pre-training on a single GPU (e.g., missing DDP data sampler) while your paper i…
-
Hi!Very good work. I have some questions. Why not consider aligning the 8 segment tokens with the generated text? would this be better
-
Hello dear authors, Thank you for your interesting work. Are you going to release the code for GCL? I am trying to experiment on different contrastive learning and I want to try you proposed GCL but I…
-
According to the paper, ULIP uses cross-modal contrastive loss to align point cloud features and image/text representations.
However, after reading the code, it seems you deploy cross entropy loss…
-
# Background
This document serves as a plan for the final publication in my PhD. This work is a continuation of the novelty-based detection scheme proposed in the last two publication. Here we discus…
-
Thank you for sharing, I refer [BalancedMSE](https://github.com/jiawei-ren/BalancedMSE)/synthetic_benchmark/loss.py;
Modified loss function in distributed run regression, after training the result is…
-
-
https://github.com/mlfoundations/open_clip/blob/37b729bc69068daa7e860fb7dbcf1ef1d03a4185/src/open_clip/loss.py#L49
In the implementation of open_clip, logits distributed across multiple gpus are …
-
### System Info
I'm getting this issue with 2x A100s for inference (or training too), but I don't get the issue with 4x A6000s.
*Environment*
```
- `transformers` version: 4.36.0.dev0
- Platfor…