Skip to content

Pull requests: unslothai/unsloth-zoo

New pull request New
73 Open 453 Closed
73 Open 453 Closed
Author
Filter by author
Loading
Label
Filter by label
Loading
Use alt + click/return to exclude labels
or ⇧ + click/return for logical OR
Projects
Filter by project
Loading
Milestones
Filter by milestone
Loading
Reviews
Filter by reviews
No reviews Review required Approved review Changes requested
Assignee
Filter by who’s assigned
Assigned to nobody Loading
Sort
Sort by
Newest Oldest Most commented Least commented Recently updated Least recently updated Best match
Most reactions
👍 👎 😄 🎉 😕 ❤️ 🚀 👀

Pull requests list

Fix CSM depth decoder generate: preserve forward signature on wrapper
#590 opened Apr 10, 2026 by danielhanchen Contributor Loading…
5 tasks done
3
Fix for Windows file locking (os error 1224)
#589 opened Apr 10, 2026 by PantelisAndrianakis Loading…
9
[Qwen 3.5] Qwen35 fast inference
#588 opened Apr 10, 2026 by Datta0 Collaborator • Draft
3
Fix inf grad_norm on Qwen3.5 at seq_len > 65536 with tighter SDPA guards
#587 opened Apr 9, 2026 by danielhanchen Contributor Loading…
3 of 5 tasks
9
Fix inf grad_norm on Qwen3.5 at seq_len > 65536 without flash-attn
#582 opened Apr 9, 2026 by danielhanchen Contributor Loading…
5 of 6 tasks
1
11
Add last_response_only parameter to train_on_responses_only
#579 opened Apr 7, 2026 by maximedb Loading…
20
Add GraniteMoeHybridForCausalLM compiler support
#562 opened Mar 24, 2026 by Maxusmusti Loading…
2 tasks done
2
Fix bugs in FP8 MoE support
#554 opened Mar 17, 2026 by danielhanchen Contributor Loading…
5 tasks
14
Fix FP8 MoE scale patching for compressed-tensors models
#551 opened Mar 16, 2026 by danielhanchen Contributor Loading…
18
Fix dead-code VLM layer count branches and missing state dict exclusion
#550 opened Mar 16, 2026 by danielhanchen Contributor Loading…
6 tasks done
3
[MoE] FP8 support for MoE, specifically GLM 4.7 flash
#548 opened Mar 16, 2026 by Datta0 Collaborator Loading…
25
Add Idefics3 fast_inference support
#540 opened Mar 12, 2026 by danielhanchen Contributor Loading…
6 tasks done
1
Double-buffer GPU activations for overlapping H2D copy with backward compute
#534 opened Mar 6, 2026 by ichbinhandsome Contributor Loading…
48
Fix _get_vllm_state_dict for LFM2 models
#531 opened Mar 3, 2026 by danielhanchen Contributor Loading…
4 tasks done
6
Moe kernels refactor
#529 opened Mar 3, 2026 by Datta0 Collaborator Loading…
7
Add Bnb4bit support for MoE models on transformers v5 - #4032
#527 opened Mar 2, 2026 by sensai99 Loading…
21
Guard GPT-OSS allocator warmup on low-memory 4-bit loads
#521 opened Feb 26, 2026 by danielhanchen Contributor Loading…
30
Fix vLLM vision GRPO compatibility for issue #4081
#520 opened Feb 26, 2026 by danielhanchen Contributor Loading…
33
Fix missing ParameterModule export in GPT-OSS compiler path
#519 opened Feb 25, 2026 by danielhanchen Contributor Loading…
9
Enable ROCm GPU acceleration for llama.cpp GGUF export
#512 opened Feb 24, 2026 by GoldenGrapeGentleman Contributor Loading…
13
Fix transformers 5.x compat: GRPO token_type_ids, gpt_oss BlockMask, compiler decorators
#511 opened Feb 24, 2026 by danielhanchen Contributor Loading…
5 tasks done
7
fix: skip non-attention layers in _get_vllm_state_dict (fixes unslothai/unsloth#4073)
#510 opened Feb 23, 2026 by stakeswky Loading…
1
4
fix: handle LFM2/Mamba hybrid layers in _get_vllm_state_dict for fast_inference
#504 opened Feb 18, 2026 by devchilll Loading…
41
Fix MoE target_parameters module_count alignment (#3405, #3701)
#499 opened Feb 14, 2026 by GoldenGrapeGentleman Contributor Loading…
15
Handle missing CSM depth decoder loss during loss aggregation
#496 opened Feb 11, 2026 by danielhanchen Contributor Loading…
10
ProTip! Add no:assignee to see everything that’s not assigned.