adjust llm readme #8672
Merged
adjust llm readme #8672
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
Thanks for your contribution! |
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## develop #8672 +/- ##
========================================
Coverage 55.79% 55.79%
========================================
Files 620 620
Lines 96659 96659
========================================
Hits 53928 53928
Misses 42731 42731 ☔ View full report in Codecov by Sentry. |
gongel
previously approved these changes
Jun 27, 2024
ZeyuChen
reviewed
Jun 27, 2024
llm/README.md
Outdated
|
|
||
| - **大模型无损量化**。大模型套件内置了PaddleSlim团队自研的自适应Shift-SmoothQuant的A8W8量化算法和业界主流GPTQ的W4量化算法，实现了主流大模型的无损量化，有效加速模型推理。 | ||
| - **大模型无损量化**。大模型套件内置了PaddleSlim团队自研的自适应Shift-SmoothQuant的A8W8量化算法和业界主流GPTQ、AWQ的W4量化算法，实现了主流大模型的无损量化，有效加速模型推理。 |
There was a problem hiding this comment.
大模型套件内置了PaddleSlim团队自研的自适应Shift-SmoothQuant的A8W8
=》大模型套件预置了PaddleSlim PTQ.LLM量化算法和业界主流GPTQ、AWQ的W4量化
| @@ -152,8 +152,8 @@ python run_finetune.py ./config/llama/pt_argument.json | |||
| 更多大模型精调分布式使用文档、训练细节和效果请参见[大模型精调教程](./docs/finetune.md)。 | |||
|
|
|||
| ### 3. 对齐 | |||
| 我们支持DPO等偏好对齐策略。DPO策略采用zero_padding策略，结合FlashMask策略，有效提升模型训练效率。 | |||
|
|
|||
| 我们支持DPO、RLHF等偏好对齐策略。DPO策略采用zero_padding策略，结合FlashMask策略，有效提升模型训练效率。 | |||
ZeyuChen
reviewed
Jun 27, 2024
llm/README.md
Outdated
| | [Bloom](./config/bloom) | ❌ | ✅ | ✅ | ✅ | 🚧 | ✅ | ✅ | | ||
| | [GPT-3](./config/gpt-3) | ✅ | ✅ | 🚧 | 🚧 | 🚧 | 🚧 | ✅ | | ||
| | [OPT](./config/opt) | 🚧 | ✅ | ✅ | 🚧 | 🚧 | 🚧 | ✅ | | ||
| | Model | Pretrain | SFT | LoRA | Prefix Tuning | DPO | RLHF | Quantization | Weight convert | |
llm/docs/rlhf.md
Outdated
| @@ -239,7 +239,7 @@ PYTHONPATH=../../ GLOG_minloglevel=2 python -u -m paddle.distributed.launch run_ | |||
|
|
|||
| ### 推理 | |||
|
|
|||
| 训练完成后可以直接使用 `outpt_dir` 所指定目录中 policy 文件夹下的 checkpoints 按照[LLM 推理](https://github.com/PaddlePaddle/PaddleNLP/tree/develop/llm#4-%E6%8E%A8%E7%90%86)部分的介绍来进行推理，请参考相应部分内容。 | |||
| 训练完成后可以直接使用 `outpt_dir` 所指定目录中 policy 文件夹下的 checkpoints 按照[LLM 推理](inference.md)部分的介绍来进行推理，请参考相应部分内容。 | |||
ZeyuChen
reviewed
Jun 27, 2024
llm/README.md
Outdated
| @@ -10,26 +10,26 @@ | |||
|
|
|||
| - **飞桨4D并行分布式策略**。 PaddleNLP Trainer 封装支持飞桨4D并行配置（数据并行、张量并行、流水线并行、 分组参数切分并行），屏蔽多硬件编程复杂性，用户可以修改Trainer配置组合多种预训练或精调过程的分布式策略，充分组合大模型4D并行训练能力，能有效提升在多模型、多硬件下的训练性能。 | |||
|
|
|||
| - **高效精调策略**。飞桨大模型套件提供SFT、PEFT等多种精调策略，搭载自研Zero Padding零填充优化策略有效减少训练数据中pad token的占比，提高模型训练效率。独创PEFT结合低比特和分布式并行策略，大幅降低大模型精调硬件门槛。 | |||
| - **高效精调对齐策略**。飞桨大模型套件提供SFT、DPO、RLHF等精调对齐策略，搭载自研Zero Padding零填充优化策略有效减少训练数据中pad token的占比，提高模型训练效率。独创PEFT结合低比特和分布式并行策略，大幅降低大模型精调硬件门槛。 | |||
There was a problem hiding this comment.
搭载这个动词一般是针对硬件？搭载XX策略比较少见，可以让大模型润色检查下。
gongel
approved these changes
Jun 27, 2024
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
PR types
Bug fixes
PR changes
Docs
Description
调整llm readme