-
As it is the only real-world dataset I currently know of, please take a look at the PROCAT dataset. It is structured in real-world data and synthetic data. Maybe you can summarize the infos in a short…
-
微博内容精选
-
Currently, various indexing tasks auto generate the segment version which is used by the overshadowing logic.
We have an use case (for Parallel index task and Local Index task) where the overshadowin…
-
@AnneSchoenauer [said](https://2investinginitiative.slack.com/archives/C050VAQACC9/p1683910479957129?thread_ts=1683893746.777999&cid=C050VAQACC9)
> If you look at this sheet https://docs.google.com…
-
**Describe the bug**
Using RoPE embeddings lead to NCCL error when training on 2 GPUs or more.
Bug was introduced in this commit: https://github.com/NVIDIA/Megatron-LM/commit/0c2074e2bdfca3a2a1ad595…
-
Hello,
Probably a trivial question:
The fine-tuning does not take a `batch_size`. It looks like input datasets are somehow grouped. Is there any best practice to decide a proper epoch num for fin…
-
While traffic_sign_code contains all the relevant info, the field with the sign description only contains the first sign.
-
This came up in the video chat discussion today and @zoq suggested that I write it down, which is definitely a good idea, to gather comments and thoughts about the idea.
When we originally made the…
-
We should revisit the design of our subsets & fields:
- Some of the main feedback we get is that the subsets and fields design is too complex and hard to understand
- It's not clear how fields shoul…
-