Tencent / Tencent/digitalhuman

The alignment loss converges very fast

Open
#7 1 comment 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
Python
Stars
361
Forks
52
PR merge metrics
No merged PRs in 30d

Description

Hi, I have tried reimplementing the alignment loss recently. The alignment loss seemed to converge very fast. After 200 steps, the alignment loss decreased to 0.02, while the cross entrope loss was 1.0 around. Is that a normal phenomenon?
Look forward to your kind reply, thank you~

Contributor guide

No contributing guide indexed for this repository

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start by reproducing the reported training behavior and inspect the alignment-loss and cross-entropy calculations used by the training entry point. Compare both loss curves around step 200 and verify whether the reported values are expected; done means documenting the cause and confirming whether a code change is required.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
machine-learning
Issue type
Bug
Difficulty
4/5
Estimated time
3-5 days
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
25/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.