WebSafe 3.7github.com
|
|
🏠
Skip to content

Pull requests: unslothai/unsloth

New pull request New
106 Open 893 Closed
106 Open 893 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 batched padded generation
#4083 opened Feb 20, 2026 by mmathew23 Loading…
1
6
Fix DDP "marked ready twice" for VLMs with CPU offload + TiledMLP
#4077 opened Feb 17, 2026 by nepfaff Loading…
9
Fix tool calling compatibility for Llama 3.2 and Phi-4
#4038 opened Feb 12, 2026 by VedantMadane Loading…
1
6
fix: use env-only SM100 workaround for vLLM PDL/MMA path
#4035 opened Feb 11, 2026 by danielhanchen Loading…
5
Guard TRL estimate_tokens warnings_issued writes in patched RL trainers
#4034 opened Feb 11, 2026 by danielhanchen Loading…
5
Fix Gemma3 NaN losses on ROCm by disabling torch.compile for RDNA GPUs
#4029 opened Feb 11, 2026 by GoldenGrapeGentleman Loading…
1
5
Fix global dequantize buffer dtype mismatch across mixed-precision loads
#4026 opened Feb 11, 2026 by GoldenGrapeGentleman Loading…
1
ROCm: default GPT-OSS to BF16 and disable AITER
#4021 opened Feb 10, 2026 by danielhanchen Loading…
17
Make bitsandbytes optional on ROCm and add bf16 helper
#4000 opened Feb 8, 2026 by danielhanchen Loading…
1
Extend TRL experimental patching and vLLM readiness
#3984 opened Feb 5, 2026 by danielhanchen Loading…
6
Fix TRL 0.25.1+ GRPO vision crash and reward function TypeError
#3975 opened Feb 3, 2026 by danielhanchen Loading…
5 tasks done
6
Add vLLM fallback and GRPO completion normalization
#3958 opened Feb 2, 2026 by danielhanchen Loading…
10
[WIP] Apple Silicon (MPS/Metal) Support
#3950 opened Jan 29, 2026 by Wilbatronic • Draft
5
[vllm] dynamically switch lora request
#3933 opened Jan 26, 2026 by Datta0 Loading…
10
Asft plus
#3918 opened Jan 21, 2026 by hcsolakoglu • Draft
4
Allow fp8 for non fast inference
#3904 opened Jan 19, 2026 by Datta0 Loading…
4
Add vLLM‑style Runtime Metrics (Inference + Training) with Opt‑In Telemetry
#3897 opened Jan 16, 2026 by hnxnq7 Loading…
10
feat: Native CGGR support for SFTTrainer (closes #3884)
#3891 opened Jan 14, 2026 by Wilbatronic Loading…
8
introduce device_context to simplify code.
#3875 opened Jan 11, 2026 by ykaitao Loading…
29
feat: add mlx model and trainer
#3856 opened Jan 6, 2026 by JINO-ROHIT Loading…
19
Add context parallelism support (SDPA only)
#3823 opened Jan 2, 2026 by djsaunde Loading…
1 of 2 tasks
@djsaunde
38
Add Qwen3-Omni Support with Optimized MTP Fine-Tuning
#3819 opened Jan 2, 2026 by AlanPonnachan Loading…
1
14
fix: propagate revision to vLLM fast_inference
#3816 opened Jan 2, 2026 by majiayu000 Loading…
6
fix: handle zero-strided tensors in fast_rope_embedding (#3781)
#3804 opened Dec 29, 2025 by yurekami Loading…
3 tasks
1
4
fix: compute position_ids correctly for batched left-padded generation
#3803 opened Dec 29, 2025 by yurekami Loading…
4 tasks
1
5
ProTip! Mix and match filters to narrow down what you’re looking for.