Skip to content

[DOC]: update documents of colossal chat #4394

@cwher

Description

@cwher

Motivation

  • Several issues regarding datasets and OOM have been proposed.
  • The code has been refactored, and we should make the documentation align with the code.

Expected Results

  • Run all examples in README. Make sure all the code snippets are runnable or we should mark them as outdated.
  • Fixed factual errors or missing parts in README, e.g., where to find the dataset?
  • Add notation for the results that are not easily reproducible, e.g., run examples that require 23.5GB device memory in a single RTX3090Ti.

Details

The following files are revised,

./examples/community/peft/README.md
./examples/community/ray/README.md
./examples/community/README.md
./examples/README.md
./benchmarks/README.md
./coati/ray/README.md
./README.md

except

./evaluate/README.md
./inference/README.md

Possible Future Work

  • Revise benchmark results since the code has been refactored.
  • Revise some algorithmic concepts. (This part is likely to be beyond my knowledge as a system engineer.)
  • Add hyper-parameter tuning tutorial. (This part is also beyond my knowledge.)

Metadata

Metadata

Assignees

Labels

No labels
No labels

Type

No type

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions