Fork
0
代码
介绍
代码
Issues
Pull Requests
流水线
Actions
讨论
Wiki
项目成员
分析
项目设置
Fork
0
master
MindSpeed-RL
/
configs
下载当前目录
ascend-robot
[fix] 更新文件 rl_config.py解决master分支32B模型劣化问题
ab6a0d16
创建于
3月23日
历史提交
文件
最后提交记录
最后更新时间
checkpoint
update core0.12.1 Co-authored-by: daikang_123<daikang6@huawei.com> # message auto-generated for no-merge-commit merge:
!671
merge master into master update core0.12.1 Created-by: daikang123 Commit-by: daikang123;daikang_123 Merged-by: ascend-robot Description: 长跑其他模型的验证结果:
https://wiki.huawei.com/domains/6995/wiki/8/WIKI202510148565350
See merge request: Ascend/MindSpeed-RL
!671
7 个月前
datasets
master 文档修改 Co-authored-by: xiazhahe<nieshiyu1@huawei.com> # message auto-generated for no-merge-commit merge:
!726
merge master into master master 文档修改 Created-by: xiazhahe Commit-by: xiazhahe Merged-by: ascend-robot Description: master 文档修改 See merge request: Ascend/MindSpeed-RL
!726
6 个月前
envs
【fix】更新readme的torch版本 Co-authored-by: xiecheng_1<xiecheng22@h-partners.com> # message auto-generated for no-merge-commit merge:
!845
merge fix_readme into master 【fix】更新readme的torch版本 Created-by: xiecheng_1 Commit-by: xiecheng_1 Merged-by: ascend-robot Description: update pytorch version in readme See merge request: Ascend/MindSpeed-RL
!845
5 个月前
model
支持search tool Co-authored-by: pengnuoheng<pengnuoheng@huawei.com> # message auto-generated for no-merge-commit merge:
!846
merge master into master 支持search tool Created-by: pengnuoheng Commit-by: pengnuoheng Merged-by: ascend-robot Description: 支持search tool 自验报告:
https://wiki.huawei.com/domains/127887/wiki/246533/WIKI202512179456276
See merge request: Ascend/MindSpeed-RL
!846
5 个月前
tools
支持search tool Co-authored-by: pengnuoheng<pengnuoheng@huawei.com> # message auto-generated for no-merge-commit merge:
!846
merge master into master 支持search tool Created-by: pengnuoheng Commit-by: pengnuoheng Merged-by: ascend-robot Description: 支持search tool 自验报告:
https://wiki.huawei.com/domains/127887/wiki/246533/WIKI202512179456276
See merge request: Ascend/MindSpeed-RL
!846
5 个月前
dapo_qwen25_32b_A2_20k.yaml
修改qwen25-32b的长序列精度问题、以及推20k随机步数oom Co-authored-by: d00613215<daikang6@huawei.com> # message auto-generated for no-merge-commit merge:
!894
merge master into master 修改qwen25-32b的长序列精度问题、以及推20k随机步数oom Created-by: daikang123 Commit-by: d00613215 Merged-by: ascend-robot Description: 修改qwen25-32b推20k,随机步数后oom 修改qwen25-32b的长序列精度问题 See merge request: Ascend/MindSpeed-RL
!894
4 个月前
dapo_qwen25_32b_A3.yaml
!484
【DAPO】filter_groups_metric参数优化 Merge pull request
!484
from pengnuoheng/master
10 个月前
dapo_qwen25_32b_A3_32k.yaml
修改qwen25-32b的长序列精度问题、以及推20k随机步数oom Co-authored-by: d00613215<daikang6@huawei.com> # message auto-generated for no-merge-commit merge:
!894
merge master into master 修改qwen25-32b的长序列精度问题、以及推20k随机步数oom Created-by: daikang123 Commit-by: d00613215 Merged-by: ascend-robot Description: 修改qwen25-32b推20k,随机步数后oom 修改qwen25-32b的长序列精度问题 See merge request: Ascend/MindSpeed-RL
!894
4 个月前
dapo_qwen25_7b_A2_multi_turn.yaml
支持search tool Co-authored-by: pengnuoheng<pengnuoheng@huawei.com> # message auto-generated for no-merge-commit merge:
!846
merge master into master 支持search tool Created-by: pengnuoheng Commit-by: pengnuoheng Merged-by: ascend-robot Description: 支持search tool 自验报告:
https://wiki.huawei.com/domains/127887/wiki/246533/WIKI202512179456276
See merge request: Ascend/MindSpeed-RL
!846
5 个月前
dapo_qwen3_30b_a3b_A3.yaml
fix cp bug Co-authored-by: fh_188<fenghui32@huawei.com> # message auto-generated for no-merge-commit merge:
!780
merge master_cp into master fix cp bug Created-by: weixin_44917616 Commit-by: fh_188 Merged-by: ascend-robot Description: 对应ms-llm修改,fix cp bug reset_position_ids替换成reset_attention_mask variable_seq_lengths替换成 no_pad_to_seq_lengths 结果对比
https://wiki.huawei.com/domains/142209/wiki/306738/WIKI202512029261035
See merge request: Ascend/MindSpeed-RL
!780
5 个月前
dapo_qwen3_32b_A3.yaml
fix cp bug Co-authored-by: fh_188<fenghui32@huawei.com> # message auto-generated for no-merge-commit merge:
!780
merge master_cp into master fix cp bug Created-by: weixin_44917616 Commit-by: fh_188 Merged-by: ascend-robot Description: 对应ms-llm修改,fix cp bug reset_position_ids替换成reset_attention_mask variable_seq_lengths替换成 no_pad_to_seq_lengths 结果对比
https://wiki.huawei.com/domains/142209/wiki/306738/WIKI202512029261035
See merge request: Ascend/MindSpeed-RL
!780
5 个月前
dpo_qwen3_30b_a3b_A3.yaml
fix cp bug Co-authored-by: fh_188<fenghui32@huawei.com> # message auto-generated for no-merge-commit merge:
!780
merge master_cp into master fix cp bug Created-by: weixin_44917616 Commit-by: fh_188 Merged-by: ascend-robot Description: 对应ms-llm修改,fix cp bug reset_position_ids替换成reset_attention_mask variable_seq_lengths替换成 no_pad_to_seq_lengths 结果对比
https://wiki.huawei.com/domains/142209/wiki/306738/WIKI202512029261035
See merge request: Ascend/MindSpeed-RL
!780
5 个月前
grpo_deepseek_r1_671b_A2.yaml
fix cp bug Co-authored-by: fh_188<fenghui32@huawei.com> # message auto-generated for no-merge-commit merge:
!780
merge master_cp into master fix cp bug Created-by: weixin_44917616 Commit-by: fh_188 Merged-by: ascend-robot Description: 对应ms-llm修改,fix cp bug reset_position_ids替换成reset_attention_mask variable_seq_lengths替换成 no_pad_to_seq_lengths 结果对比
https://wiki.huawei.com/domains/142209/wiki/306738/WIKI202512029261035
See merge request: Ascend/MindSpeed-RL
!780
5 个月前
grpo_deepseek_r1_671b_A3.yaml
fix cp bug Co-authored-by: fh_188<fenghui32@huawei.com> # message auto-generated for no-merge-commit merge:
!780
merge master_cp into master fix cp bug Created-by: weixin_44917616 Commit-by: fh_188 Merged-by: ascend-robot Description: 对应ms-llm修改,fix cp bug reset_position_ids替换成reset_attention_mask variable_seq_lengths替换成 no_pad_to_seq_lengths 结果对比
https://wiki.huawei.com/domains/142209/wiki/306738/WIKI202512029261035
See merge request: Ascend/MindSpeed-RL
!780
5 个月前
grpo_deepseek_r1_671b_A3_eplb.yaml
fix cp bug Co-authored-by: fh_188<fenghui32@huawei.com> # message auto-generated for no-merge-commit merge:
!780
merge master_cp into master fix cp bug Created-by: weixin_44917616 Commit-by: fh_188 Merged-by: ascend-robot Description: 对应ms-llm修改,fix cp bug reset_position_ids替换成reset_attention_mask variable_seq_lengths替换成 no_pad_to_seq_lengths 结果对比
https://wiki.huawei.com/domains/142209/wiki/306738/WIKI202512029261035
See merge request: Ascend/MindSpeed-RL
!780
5 个月前
grpo_lora_qwen25_32b_A2.yaml
fix cp bug Co-authored-by: fh_188<fenghui32@huawei.com> # message auto-generated for no-merge-commit merge:
!780
merge master_cp into master fix cp bug Created-by: weixin_44917616 Commit-by: fh_188 Merged-by: ascend-robot Description: 对应ms-llm修改,fix cp bug reset_position_ids替换成reset_attention_mask variable_seq_lengths替换成 no_pad_to_seq_lengths 结果对比
https://wiki.huawei.com/domains/142209/wiki/306738/WIKI202512029261035
See merge request: Ascend/MindSpeed-RL
!780
5 个月前
grpo_qwen25_32b_A2.yaml
gptoss patch 原仓合入 Co-authored-by: nuerxiati<738457498@qq.com> # message auto-generated for no-merge-commit merge:
!910
merge update_gptoss into master gptoss patch 原仓合入 Created-by: NurxatAbilmit Commit-by: NurxatAbilmit;nuerxiati Merged-by: ascend-robot Description: 此前vllm-ascend 因为sink功能未合入,使用patch方式运行,当前代码以合入到vllm-ascend主线分支 See merge request: Ascend/MindSpeed-RL
!910
2 个月前
grpo_qwen25_32b_A3.yaml
[fix] 更新文件 rl_config.py解决master分支32B模型劣化问题 Co-authored-by: wangshuyang31<wangshuyang8@huawei.com> # message auto-generated for no-merge-commit merge:
!920
merge master into master [fix] 更新文件 rl_config.py解决master分支32B模型劣化问题 Created-by: wangshuyang31 Commit-by: wangshuyang31 Merged-by: ascend-robot Description: update: 更新文件 rl_config.py See merge request: Ascend/MindSpeed-RL
!920
2 个月前
grpo_qwen25_7b_A3.yaml
gptoss patch 原仓合入 Co-authored-by: nuerxiati<738457498@qq.com> # message auto-generated for no-merge-commit merge:
!910
merge update_gptoss into master gptoss patch 原仓合入 Created-by: NurxatAbilmit Commit-by: NurxatAbilmit;nuerxiati Merged-by: ascend-robot Description: 此前vllm-ascend 因为sink功能未合入,使用patch方式运行,当前代码以合入到vllm-ascend主线分支 See merge request: Ascend/MindSpeed-RL
!910
2 个月前
grpo_qwen3_235b_a22b_A2.yaml
fix cp bug Co-authored-by: fh_188<fenghui32@huawei.com> # message auto-generated for no-merge-commit merge:
!780
merge master_cp into master fix cp bug Created-by: weixin_44917616 Commit-by: fh_188 Merged-by: ascend-robot Description: 对应ms-llm修改,fix cp bug reset_position_ids替换成reset_attention_mask variable_seq_lengths替换成 no_pad_to_seq_lengths 结果对比
https://wiki.huawei.com/domains/142209/wiki/306738/WIKI202512029261035
See merge request: Ascend/MindSpeed-RL
!780
5 个月前
grpo_qwen3_8b_A3.yaml
!586
fix CP + removing padding Merge pull request
!586
from 戴康123/master
8 个月前
ppo_qwen25_32b_A3.yaml
fix cp bug Co-authored-by: fh_188<fenghui32@huawei.com> # message auto-generated for no-merge-commit merge:
!780
merge master_cp into master fix cp bug Created-by: weixin_44917616 Commit-by: fh_188 Merged-by: ascend-robot Description: 对应ms-llm修改,fix cp bug reset_position_ids替换成reset_attention_mask variable_seq_lengths替换成 no_pad_to_seq_lengths 结果对比
https://wiki.huawei.com/domains/142209/wiki/306738/WIKI202512029261035
See merge request: Ascend/MindSpeed-RL
!780
5 个月前