Skip to content

Conversation

@liym27
Copy link
Contributor

@liym27 liym27 commented Mar 20, 2024

PR types

New features

PR changes

Models

Description

support GPT for auto_parallel

1. 在GPT上验证自动并行架构

  • 对比动态图半自动并行和动态图手动并行,收敛和精度结果符合预期,涉及以下策略验证
    • mp2
    • dp2mp2
    • dp2mp2pp2
    • dp2mp2pp2 + 动转静
    • dp2mp2pp2 + 动转静 + amp

    dp 数据并行,mp 模型并行,pp 流水线并行, amp 自动混合精度

2. 本 PR 代码包含最终模型代码、示例脚本、单测。验证结果如下

mp2

image

dp2mp2

image

dp2mp2pp2

image

dp2mp2pp2 + 动转静

image

dp2mp2pp2 + 动转静 + amp

image image
@paddle-bot
Copy link

paddle-bot bot commented Mar 20, 2024

Thanks for your contribution!

@codecov
Copy link

codecov bot commented Mar 20, 2024

Codecov Report

Attention: Patch coverage is 16.50672% with 435 lines in your changes are missing coverage. Please review.

Project coverage is 54.88%. Comparing base (7b493a8) to head (f368f57).
Report is 2 commits behind head on develop.

❗ Current head f368f57 differs from pull request most recent head 1b0f2ef. Consider uploading reports for the commit 1b0f2ef to get more accurate results

Files Patch % Lines
paddlenlp/transformers/gpt/modeling_auto.py 16.34% 435 Missing ⚠️
Additional details and impacted files
@@ Coverage Diff @@ ## develop #8160 +/- ## =========================================== - Coverage 55.15% 54.88% -0.27%  =========================================== Files 601 603 +2 Lines 91764 92371 +607 =========================================== + Hits 50611 50699 +88  - Misses 41153 41672 +519 

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

@liym27 liym27 force-pushed the gpt_auto_parallel branch 6 times, most recently from ac9b855 to 5b6cb9d Compare March 21, 2024 06:31
@liym27 liym27 marked this pull request as draft March 21, 2024 07:24
@liym27 liym27 closed this Mar 21, 2024
@liym27 liym27 reopened this Mar 21, 2024
@liym27 liym27 marked this pull request as ready for review March 21, 2024 07:32
liym27 added 3 commits April 1, 2024 10:58
add run_pretrain modify llama test loss
comment llama
@liym27 liym27 force-pushed the gpt_auto_parallel branch 4 times, most recently from d2ef691 to d9f6b83 Compare April 2, 2024 06:19
@liym27 liym27 force-pushed the gpt_auto_parallel branch from d9f6b83 to f901267 Compare April 2, 2024 06:25
@liym27 liym27 closed this Apr 6, 2024
@liym27 liym27 reopened this Apr 6, 2024
@liym27 liym27 closed this Apr 7, 2024
@liym27 liym27 reopened this Apr 7, 2024
@liym27 liym27 force-pushed the gpt_auto_parallel branch from f368f57 to 1b0f2ef Compare April 7, 2024 14:26
@PaddlePaddle PaddlePaddle locked and limited conversation to collaborators Apr 8, 2024
@PaddlePaddle PaddlePaddle unlocked this conversation Apr 8, 2024
@PaddlePaddle PaddlePaddle locked and limited conversation to collaborators Apr 8, 2024
@PaddlePaddle PaddlePaddle unlocked this conversation Apr 8, 2024
Copy link
Contributor

@ZHUI ZHUI left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM

@wawltor wawltor merged commit 98a4b84 into PaddlePaddle:develop Apr 10, 2024
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

3 participants