lm-sys / lm-sys/FastChat

[train flant5] Preprocessing the conversations dataset is too slow.

Open
#2,546 1 comment 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
Python
Stars
39.5k
Forks
4.8k
PR merge metrics
No merged PRs in 30d

Description

train_flant5.py line236

for source in sources:
    conversation = _add_speaker_and_signal(header, source, tokenizer)
    conversations.append(conversation)

Passing the tokenizer object as a boolean argument makes the operation slow (1000X).

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 in fastchat/train/train_flant5.py around line 236 and inspect how _add_speaker_and_signal receives the tokenizer argument. Reproduce the preprocessing timing with the current call, then verify that the corrected argument handling removes the severe slowdown without changing conversation preprocessing results.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
machine-learning
Issue type
Bug
Difficulty
2/5
Estimated time
1-3 hours
Activity status
Stale
Clarity
Mostly clear
Newbie friendliness
42/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.