Commits · 535c37b6b99012c0bfaa2b644e65c84918e02637 · evt_fugx1 / dcu_megatron

16 May, 2025 1 commit
- update model parameters format · 535c37b6
  silencealiang authored May 16, 2025
  
  535c37b6
15 May, 2025 1 commit
- disable torch._dynamo for reduce_from_tensor_model_parallel_region · 83fab71e
  silencealiang authored May 15, 2025
  
  83fab71e
14 May, 2025 2 commits
- 添加reproduce功能, 开启后可实现训练loss完全复现, 少量降低训练性能 · eb4333f0
  wangxj authored May 14, 2025
  
  eb4333f0
- update model parameters format · 57944e55
  silencealiang authored May 14, 2025
  
  57944e55
12 May, 2025 2 commits
- Update run_llama2_7B_1nodes.sh · 90ae7f5c
  silencealiang authored May 12, 2025
  
  90ae7f5c
- fix llama2 bug and update file format · fe0b03b5
  silencealiang authored May 12, 2025
  
  fe0b03b5
07 May, 2025 3 commits
- Update deepseekv3 parameters to avoid expert select fault · ee3ff5df
  silencealiang authored May 07, 2025
  
  ee3ff5df
- check te version · 54740897
  dongcl authored May 07, 2025
  
  54740897
- fix flux import error · 2862a32a
  dongcl authored May 07, 2025
  
  2862a32a
06 May, 2025 3 commits
- modify TEGroupedLinear base · 23eb9b17
  dongcl authored May 06, 2025
  
  23eb9b17
- bug fix · 43770f8e
  dongcl authored May 06, 2025
  
  43770f8e
- Update train_deepseekv3_671B_multinodes.sh · b85974a6
  silencealiang authored May 06, 2025
  
  b85974a6
30 Apr, 2025 5 commits
- modify gpt_model · 84af95b0
  sdwldchl authored Apr 30, 2025
  
  84af95b0
- fix bug · bc3d72d1
  silencealiang authored Apr 30, 2025
  
  bc3d72d1
- rm __pycache__ · bcd9442e
  silencealiang authored Apr 30, 2025
  
  bcd9442e
- update · 48cf4c79
  silencealiang authored Apr 30, 2025
  
  48cf4c79
- update model parameters · 70368616
  silencealiang authored Apr 30, 2025
  
  70368616
29 Apr, 2025 5 commits
- bug fix · 8551c38e
  silencealiang authored Apr 29, 2025
  
  8551c38e
- modify Megatron-LM · bcb9d73e
  dongcl authored Apr 29, 2025
  
  bcb9d73e
- modify megatron submodule · 02676184
  dongcl authored Apr 29, 2025
  
  02676184
- modify megatron submodule · 237dfffa
  dongcl authored Apr 29, 2025
  
  237dfffa
- patch for megatron 4429e8ebe · a8a2bbea
  dongcl authored Apr 29, 2025
  
  a8a2bbea
25 Apr, 2025 2 commits
- 删除llama2_7b.sh · 2ddbd4be
  dongcl authored Apr 25, 2025
  
  2ddbd4be
- bug fix · 8b656f45
  silencealiang authored Apr 25, 2025
  
  8b656f45
24 Apr, 2025 5 commits
- checkout Megatron-LM to main · 1f73f815
  silencealiang authored Apr 24, 2025
  
  1f73f815
- checkout Megatron-LM to main · c593bb40
  silencealiang authored Apr 24, 2025
  
  c593bb40
- merge megatron_v0.11.0 · 02eacdd3
  dongcl authored Apr 24, 2025
  
  02eacdd3
- merge megatron_v0.11.0 · 1f376414
  dongcl authored Apr 24, 2025
  
  1f376414
- bug fix · c140c914
  dongcl authored Apr 24, 2025
  
  c140c914
23 Apr, 2025 2 commits
- modify rms_norm.py · 26940c4c
  dongcl authored Apr 23, 2025
  
  26940c4c
- modify ParallelAttention · 763941b5
  dongcl authored Apr 23, 2025
  
  763941b5
21 Apr, 2025 5 commits
- Merge branch 'megatron_v0.11.0' into 'main' · eefee67c
  dongcl authored Apr 21, 2025
```
Megatron v0.11.0

See merge request OpenDAS/dcu_megatron!3
```
  eefee67c
- Megatron v0.11.0 · ce02cd51
  dongcl authored Apr 21, 2025
  
  ce02cd51
- support flux for mtp · bf323343
  dongcl authored Apr 21, 2025
  
  bf323343
- Merge branch 'megatron_v0.11.0' of... · 31e933a8
  dongcl authored Apr 21, 2025
```
Merge branch 'megatron_v0.11.0' of http://developer.sourcefind.cn/codes/OpenDAS/dcu_megatron into megatron_v0.11.0
```
  31e933a8
- bug fix · bf212e29
  dongcl authored Apr 21, 2025
  
  bf212e29
18 Apr, 2025 3 commits
- check flux version · 4754c307
  dongcl authored Apr 18, 2025
  
  4754c307
- update README.md · b3371c31
  dongcl authored Apr 18, 2025
  
  b3371c31
- use flux gemmrs · 595e428a
  dongcl authored Apr 18, 2025
  
  595e428a
17 Apr, 2025 1 commit
- modify gpt_model · 0867fd90
  dongcl authored Apr 17, 2025
  
  0867fd90