deng
Public Activity Repositories Starred Repositories
84
 Following
19
 Followers
4
 Achievement Badge
1
Loading Heatmap…

deng synced commits to master at deng/mindformers from mirror

  • 7b8b27fe9a !7869 【master】【bugfix】修复sharded_tesnor和metadata测试用例 Merge pull request !7869 from 森镇/fix_test_ut_of_sharded_tensor
  • 13add2b32e !7866 【bugfix】【master】去除Qwen3系列中数据集无用配置，避免增加额外操作引起性能波动 Merge pull request !7866 from hsshuai/bugfix/master/dataset_setting
  • d3b3bbc2a3 修复sharded_tesnor和metadata测试用例
  • 9f0a67b57f !7837 修改blended_megatron_dataset_builder和gpt_data偶发错误 Merge pull request !7837 from zzzkeke/new/add_builder_test
  • 477621fde7 Update dataset configurations to remove attention_mask in Qwen3 YAML
  • Compare 6 commits »

23 hours ago

deng synced commits to master at deng/mindscience from mirror

3 days ago

deng synced commits to master at deng/mindformers from mirror

  • f2e32b6445 !7861 【master】更新三方依赖版本信息 Merge pull request !7861 from zyw_hw/update_transformers_copyright
  • 94d293ae67 update transformers copyright
  • 69ade06f59 !7743 【master】【infer】修复共享专家大于1场景下hidden_size赋值逻辑 Merge pull request !7743 from pengjingyou/bugfix_shared_experts
  • fe3e6a8fc3 【master】【infer】修复共享专家大于1场景下hidden_size赋值逻辑
  • Compare 4 commits »

3 days ago

deng synced commits to br_feature_llm_trainer at deng/mindformers from mirror

3 days ago

deng synced commits to r1.7.0 at deng/mindformers from mirror

  • 61f2d8bd53 !7854 【r1.7.0】将research中telechat模型里面的mint.split算子替换为ops.function.array_func.split_ext Merge pull request !7854 from zhouxq/split_r1.7.0
  • 40f9cdeae7 【r1.7.0】将research中telechat模型里面的mint.split算子替换为ops.function.array_func.split_ext
  • Compare 2 commits »

5 days ago

deng synced commits to master at deng/mindformers from mirror

  • b768b5c1e5 !7846 【master】权重2.0部分函数重构 Merge pull request !7846 from 森镇/fix_apply_balanced_load
  • a767dadbfb !7841 【master】【门禁】更新ms包 Merge pull request !7841 from zyw_hw/update_ms_pkg_1209
  • eb8ddefc51 update ms pkg
  • 34a6ef54c4 !7852 【master】更新三方依赖版本信息 Merge pull request !7852 from zyw_hw/update_megatron_version
  • 679e210ddd !7847 【master】【infer】新增Glm4Moe整网减层st用例 Merge pull request !7847 from pengjingyou/glm4_st
  • Compare 12 commits »

5 days ago

deng synced commits to br_feature_llm_trainer at deng/mindformers from mirror

5 days ago

deng synced commits to master at deng/mindformers from mirror

  • 517802fc69 !7851 【master】修改qk_clip_threshold默认值为100 Merge pull request !7851 from JavaZero/reset_default_qk_clip_threshold
  • 567b89272d !7833 【master】添加Muon优化器的ut Merge pull request !7833 from JavaZero/test_muon
  • 1c65f2afff update: increase default qk_clip_threshold from 4 to 100
  • 7bba670f5c test: add unit tests for Muon optimizer initialization and computation test: add baseline data and comparison for Muon optimizer losses
  • Compare 4 commits »

6 days ago

deng synced commits to master at deng/mindscience from mirror

  • 497660d230 !2449 fix a bug from the rfantibody github repository Merge pull request !2449 from chendanyang/master
  • 7c2ac216d3 fix a bug from the rfantibody github repository Signed-off-by: chendanyang <birfychen@gmail.com>
  • Compare 2 commits »

6 days ago

deng synced commits to r1.7.0 at deng/mindformers from mirror

  • 9e5735022e !7848 【r1.7.0】将mint.split算子替换为ops.function.array_func.split_ext Merge pull request !7848 from zhouxq/split_r1.7.0
  • 5c022807a5 【r1.7.0】将mint.split算子替换为ops.function.array_func.split_ext
  • Compare 2 commits »

6 days ago

deng synced commits to master at deng/mindformers from mirror

  • 2261fe1903 !7688 【master】SlidingWindowAttention和SharedCrossAttention模块新增测试用例 Merge pull request !7688 from lan/yoco_test
  • 7e350ad94f SlidingWindowAttention和SharedCrossAttention新增测试用例
  • Compare 2 commits »

6 days ago

deng synced commits to br_feature_llm_trainer at deng/mindformers from mirror

6 days ago

deng synced commits to master at deng/mindformers from mirror

  • f106c71c19 !7845 bugfix】【master】【docs】deepseek和telechat文档修改 Merge pull request !7845 from Yule100/code_docs_telechat2_infer
  • 1ec8a5cff3 bugfix 文档修复
  • 08f2e2215c !7826 【master】【bugfix】消除cp MOE模块的冗余重排 Merge pull request !7826 from lzy0920232/code_bugfix_cp_moe
  • b0a2c47564 code_bugfix_cp_moe
  • Compare 4 commits »

1 week ago

deng synced commits to master at deng/mindscience from mirror

  • c08f83e1a2 !2435 refactor_orb_v2 Merge pull request !2435 from liumintao2025/refactor_orb_v2
  • c2a52f5fe0 !2440 refactor diffcsp Merge pull request !2440 from liumintao2025/refactor_diffcsp
  • 13306ed300 !2437 refactor nequip: move mindchemistry to models Merge pull request !2437 from liumintao2025/refactor_nequip
  • 9aa42ecd21 !2438 refactor matformer Merge pull request !2438 from liumintao2025/refactor_matformer
  • f47d2919a8 !2445 【MindSPONGE】move AF3 from legacy-master to master Merge pull request !2445 from Yuheng Wang/af3_move
  • Compare 10 commits »

1 week ago

deng synced commits to r1.7.0 at deng/mindformers from mirror

  • d9fca4b2c4 !7844 【r1.7.0】【bugfix】【日志】权重相关日志，在raise Error之前添加logger.error，确保在error.log中有对应日志 Merge pull request !7844 from SaiYao/add_error_log_in_ckpt_r170
  • 8733fdbd6d 【r1.7.0】【bugfix】【日志】权重相关日志，在raise Error之前添加logger.error，确保在error.log中有对应日志
  • Compare 2 commits »

1 week ago

deng synced commits to master at deng/mindformers from mirror

  • 78628e0f6c !7824 stable_rank_fix_moe Merge pull request !7824 from 宋佳琪/stable_rank_fix_moe
  • 8b8df9de27 !7840 【master】【bugfix】【权重】离线转换脚本参数名统一，与convert_weight.py保持一致 Merge pull request !7840 from SaiYao/rename_argparse_of_convert_weight
  • 731b7197a8 stable_rank_fix_moe A
  • 660417b1fe !7843 [Docs] Delete expired statement in llama3.1 README.md Merge pull request !7843 from Xinrui Chen/code_docs-master-llama3-1
  • 0a98712617 !7803 【master】添加权重参数广播功能 Merge pull request !7803 from 森镇/add_param_broadcast
  • Compare 28 commits »

1 week ago

deng synced commits to master at deng/mindformers from mirror

  • a43d7e1540 !7825 【master】修复专家权重op未切满的场景下的校验 Merge pull request !7825 from JavaZero/fix_muon_ep_valid
  • e5bb19adea refactor: update MoE expert validation logic in Muon and GPTModel
  • Compare 2 commits »

1 week ago

deng synced commits to master at deng/mindformers from mirror

  • 62526e2d43 !7823 【master】重构max_logits监控流程，从参数分组逻辑中解耦出来 Merge pull request !7823 from JavaZero/update_max_logits
  • 739411b385 !7822 【master】【bugfix】【日志】权重相关日志，在raise Error之前添加logger.error，确保在error.log中有对应日志 Merge pull request !7822 from SaiYao/add_error_log_in_ckpt
  • 211024cc01 !7802 【master】【bugfix】【muon】add sharded_state_dict for muon op group Merge pull request !7802 from husichao/master
  • 802f3516c5 【master】【bugfix】【日志】权重相关日志，在raise Error之前添加logger.error，确保在error.log中有对应日志
  • 43fd826ae2 add sharded_state_dict for muon op group
  • Compare 14 commits »

1 week ago

deng synced commits to master at deng/mindformers from mirror

  • f271c58cd2 !7786 新增modeling_utils测试用例 Merge pull request !7786 from lan/callback_test_
  • cb3083aab9 !7810 【bugfix】【master】修复测试用例bug Merge pull request !7810 from hsshuai/test/master/fix_trainer_case
  • 656b143f71 !7807 修复test_pma用例超时，去掉test_all_reduce中无用用例 Merge pull request !7807 from lan/pma_test
  • 4e95ab6f51 !7808 修改blended_megatron_dataset_builder测试用例构建失败用例 Merge pull request !7808 from zzzkeke/new/add_builder_test
  • 1738940a3d !7800 【MindFormers】【覆盖率】mindformers用例覆盖率较低，补充用例并行解码和流式推理 Merge pull request !7800 from Yule100/inference_ut
  • Compare 42 commits »

1 week ago

deng synced commits to master at deng/mindscience from mirror

  • 8f019a6d75 !2414 【开源实习】 PowerFlowNet网络结构迁移 Merge pull request !2414 from wushuo2025/upstream
  • 9930a9f58c update CODEOWNERS. Signed-off-by: Chen_Yu <chenyu289@huawei.com>
  • 9d85f4e6b2 update CODEOWNERS. Signed-off-by: Chen_Yu <chenyu289@huawei.com>
  • 0f59f78adc update powerflownet structure and migration readme.md
  • Compare 4 commits »

1 week ago

Please read the following content carefully:

Dear OpenI User

Thank you for your continuous support to the Openl Qizhi Community AI Collaboration Platform. In order to protect your usage rights and ensure network security, we updated the Openl Qizhi Community AI Collaboration Platform Usage Agreement in January 2024. The updated agreement specifies that users are prohibited from using intranet penetration tools. After you click "Agree and continue", you can continue to use our services. Thank you for your cooperation and understanding.

For more agreement content, please refer to the《Openl Qizhi Community AI Collaboration Platform Usage Agreement》

Agree and continue
Disagree, exit
Baidu
map