-
Notifications
You must be signed in to change notification settings - Fork 39
Pull requests: modelscope/mcore-bridge
Author
Label
Projects
Milestones
Reviews
Assignee
Sort
Pull requests list
Fix two FP8-training issues: PLE ngram dequantization and LoRA fp8 autocast guard
#183
opened Sep 3, 2026 by
ehooon
Loading…
[bugfix] preserve QSA MRoPE batch dimension
#179
opened Sep 1, 2026 by
taking-lying-flat
Contributor
Loading…
feat(model): add generic chunked linear cross entropy
#177
opened Aug 31, 2026 by
maomaocun
Loading…
[model] refactor qwen3.5 GDN to use FLA causal conv1d and cu_seq_lens_q
#176
opened Aug 31, 2026 by
ys2025-AI
Contributor
Loading…
[bugfix] avoid recomputing Flash-decode RoPE cache
#160
opened Aug 6, 2026 by
taking-lying-flat
Contributor
Loading…
[bugfix] support MTP-only pipeline stages
#155
opened Jul 31, 2026 by
taking-lying-flat
Contributor
Loading…
Fix assertion error when packed_seq_params is passed without proper qkv_format
#109
opened Jun 4, 2026 by
tomzw11
Loading…
feat(kimi_k25): implement multimodal get_inputs_embeds for Megatron training
#106
opened Jun 3, 2026 by
zhihanliu-collab
•
Draft
[feat] Add initial Gemma 4 multimodal bridge supportFeat/gemma4 support
#15
opened Apr 5, 2026 by
JimmyMa99
Loading…
ProTip!
Type g i on any issue or pull request to go back to the issue listing page.