-
Notifications
You must be signed in to change notification settings - Fork 1.5k
Pull requests: jundot/omlx
Author
Label
Projects
Milestones
Reviews
Assignee
Sort
Pull requests list
fix(audio): reject unsupported TTS response_format with 400
#2231
opened Jul 14, 2026 by
JimStenstrom
Contributor
Loading…
2
fix(admin): give the accuracy bench queue chain an ownership token
#2230
opened Jul 14, 2026 by
JimStenstrom
Contributor
Loading…
fix(dflash): stop double-charging hot-cache bytes in the primary prefill guard
#2229
opened Jul 14, 2026 by
JimStenstrom
Contributor
Loading…
fix: load indexed MTP sidecar safetensors skipped by mlx-lm's glob
#2220
opened Jul 12, 2026 by
mxl
Contributor
Loading…
chore(deps): update transformers requirement from <5.13,>=5.7.0 to >=5.7.0,<5.14
dependencies
Pull requests that update a dependency file
python
Pull requests that update python code
#2214
opened Jul 12, 2026 by
dependabot
Bot
Loading…
feat: add YaRN RoPE scaling for Qwen3.5/3.6 MRoPE VLMs
#2210
opened Jul 12, 2026 by
a4501150
Contributor
Loading…
6 tasks done
feat(oq): add preserve_gdn to keep GDN/linear_attn layers at fp16
#2207
opened Jul 11, 2026 by
a4501150
Contributor
Loading…
3 tasks done
fix(oqe): collect lm_head imatrix and skip embedding lookup
#2206
opened Jul 11, 2026 by
a4501150
Contributor
Loading…
3 tasks done
fix(oq): scope VisionConfig model_type normalization to Qwen types
#2205
opened Jul 11, 2026 by
a4501150
Contributor
Loading…
3 tasks done
feat: warn at startup when installed git-pinned deps drift from pyproject
#2176
opened Jul 10, 2026 by
richgoodson
Contributor
Loading…
3 tasks done
fix: accept gemma4_unified in the DFlash compatibility gate (#2153)
#2175
opened Jul 10, 2026 by
richgoodson
Contributor
Loading…
3 tasks done
fix(mtp): unblock MTP under continuous batching — singleton recovery + opt-in row-wise batch path (#2150)
#2166
opened Jul 10, 2026 by
gilby
Contributor
Loading…
feat(oq): layer-streaming imatrix calibration for models larger than RAM
#2161
opened Jul 10, 2026 by
builtbycnob
Contributor
Loading…
feat(vlm): DeepSeek-V4 via VLM engine — enable vlm_mtp + DSML tool-calling
#2090
opened Jul 4, 2026 by
farolone
Loading…
fix(vlm): torch-free AutoProcessor fallback for gated processor classes
#2088
opened Jul 4, 2026 by
tbro0815
Contributor
Loading…
fix(server): defer non-streaming JSON keepalive to the first interval
#2072
opened Jul 2, 2026 by
JimStenstrom
Contributor
Loading…
feat: add adapter_path support for LoRA fine-tuned models
#2047
opened Jun 30, 2026 by
sje397-automation
Contributor
Loading…
[codex] update Russian admin localization
#2027
opened Jun 28, 2026 by
DrMaks22
Contributor
Loading…
fix(api): generate Mistral-compliant tool-call ids
#1997
opened Jun 24, 2026 by
sam-fakhreddine
Loading…
fix(engine): enable tool calling for MistralCommon (tekken) tokenizers
#1996
opened Jun 24, 2026 by
sam-fakhreddine
Loading…
fix: show DFlash prefill/generation status in admin dashboard
#1993
opened Jun 24, 2026 by
dwojcik92
Loading…
Previous Next
ProTip!
Updated in the last three days: updated:>2026-07-11.