admin/LLM-from-scratch
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

[25.05.26] add documents->tokenizer

Browse Source
This commit is contained in:
hanluzhi0429@gmail.com
2025-05-26 22:35:30 +08:00
parent 87fc8933ec
commit 61b8b42b3c
1 changed files with 3 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files
README.md
+3
View File
@@ -4,6 +4,9 @@ LLM-Dolly is a custom LLM built from scratch
## 🗓️ 更新日志 ## 🗓️ 更新日志
#### 2025-05-26
- 📝 增加LLM文档，涵盖LLM全周期技术点。
#### 2025-05-22 #### 2025-05-22
- 📝 添加支持MoE模型，训练资源占用不稳定，测试模型Experts=8, experts_per_tok=4,GPU显存从60%-94%跳动,单个GPU利用率0%-100%跳动。当调小batch_size后，显存占用稳定在90%，单个GPU利用率稳定在90%以上，偶现30%的利用率。 - 📝 添加支持MoE模型，训练资源占用不稳定，测试模型Experts=8, experts_per_tok=4,GPU显存从60%-94%跳动,单个GPU利用率0%-100%跳动。当调小batch_size后，显存占用稳定在90%，单个GPU利用率稳定在90%以上，偶现30%的利用率。