Who contributed to Hugging Face on July 15, 2026?

13 developers shipped this update, including remi-or, David-Wu1119, Harry Mellor, kaixuanliu, dacorvo, sywangyi, Abdennacer-Badaoui, and vasqu, and 5 more.

What were the notable Hugging Face updates?

Fix half-precision torch.compile crash in DETR-family sine position embeddings, aiter-rope: Add layer repo for apply_rotary_transformers, and liger-kernels: Add layer repos for remaining functions.

@huggingface

Transformers, Datasets, and the open AI-model layer

github ↗

Pick a date

Topics: Python AI / ML Full archive →

The Wire · Showcase

TRANSFORMERS PATCHES TORCH.COMPILE CRASH; KERNELS SHIP LAYER REGISTRY UPDATES

By RepoJournal · Filed 07:00 UTC on July 15, 2026 · About Hugging Face

13 people shipped this

David-Wu1119 @David-Wu1119 1 cited

Harry Mellor @hmellor 1 cited

sywangyi @sywangyi 1 cited

Abdennacer-Badaoui @Abdennacer-Badaoui 1 cited

vasqu @vasqu 1 cited

stevhliu @stevhliu 1 cited

yzhautouskay @yzhautouskay 1 cited

Arek Borucki @arekborucki 1 cited

remi-or @remi-or

kaixuanliu @kaixuanliu

dacorvo @dacorvo

menglcai @menglcai

YiYi Xu @yiyixuxu

Transformers just fixed a half-precision torch.compile crash that broke DETR models in v5.9.0, while the kernel ecosystem shipped three layer-repo updates that unify the custom kernel interface across rotary, liger, and rope implementations.

In brief

Transformers patched a half-precision torch.compile crash in DETR sine position embeddings. One less footgun when quantizing vision models.
GPTBigCode and GPTNeoX now work with vLLM's Transformers modelling backend. Inference serving just got less fragmented.
Kernels-community added layer repos for aiter-rope (apply_rotary_transformers). The kernel registry is getting clearer.
Liger-kernels filled out remaining layer repos. Kernel composition is leveling up.
finegrained-fp8 shipped a GRF register-pressure pruner for XPU autotune. Squeezing more efficiency out of specialized hardware.

The torch.compile regression [1] hit DETR position embeddings when a recent rewrite of sine embeddings triggered Inductor's pointless_cumsum_replacement pattern, causing dtype mismatches that crashed encoder attention under bf16. This is a direct regression from v5.8.0 and needs immediate patches for anyone compiling DETR variants. Meanwhile, kernels-community shipped three coordinated PRs [2] [3] that add nn.Module layer versions for apply_rotary_transformers, remaining liger functions, and rotary embeddings respectively, mirroring the layer-repo pattern in transformers#46883. These standardize how custom kernels expose both functional and module interfaces, making them compatible with transformers' layer-registry loading path. On the vLLM front, transformers fixed GPTBigCode and GPTNeoX for the new modelling backend [4], and the kernels team landed a smart XPU autotune improvement [5] that prunes high-register-pressure configs before compilation to skip tens of thousands of spilled instructions and wasted compile cycles. Diffusers added torch.dtype aliasing for Hub snippet compatibility [6] and shipped Cosmos3 transfer controls [7] for structure-conditioned video generation. Dataset-viewer merged IRSA support [8] for prod datasets-server, moving away from static S3 keys.

Action items

→ Apply transformers torch.compile fix [ref:2] if you're compiling DETR models in bf16 huggingface/transformers [immediate]
→ Review and integrate the three kernel layer-repo updates [ref:10] [ref:11] if you maintain custom kernel wrappers huggingface/kernels-community [plan]
→ Monitor vLLM GPTBigCode/GPTNeoX serving [ref:3] after the modelling backend fix lands huggingface/transformers [monitor]

References

[1] Fix half-precision torch.compile crash in DETR-family sine position embeddings ↗ huggingface/transformers
[2] aiter-rope: Add layer repo for apply_rotary_transformers ↗ huggingface/kernels-community
[3] liger-kernels: Add layer repos for remaining functions ↗ huggingface/kernels-community
[4] :rotating_light: Fix GPTBigCode and GPTNeoX for the Transformers modelling backend for vLLM (#47198) huggingface/transformers
[5] finegrained-fp8: add GRF register-pressure pruner for XPU autotune ↗ huggingface/kernels-community
[6] torch.dtype alias ↗ huggingface/diffusers
[7] Transfer support for Cosmos3 ModularPipeline ↗ huggingface/diffusers
[8] Add IRSA to datasets-server (#3359) huggingface/dataset-viewer

Quick answers

What shipped in Hugging Face on July 15, 2026?: Transformers just fixed a half-precision torch.compile crash that broke DETR models in v5.9.0, while the kernel ecosystem shipped three layer-repo updates that unify the custom kernel interface across rotary, liger, and rope implementations. In total, 20 commits and 19 pull requests landed.
Who contributed to Hugging Face on July 15, 2026?: 13 developers shipped this update, including remi-or, David-Wu1119, Harry Mellor, kaixuanliu, dacorvo, sywangyi, Abdennacer-Badaoui, and vasqu, and 5 more.
What were the notable Hugging Face updates?: Fix half-precision torch.compile crash in DETR-family sine position embeddings, aiter-rope: Add layer repo for apply_rotary_transformers, and liger-kernels: Add layer repos for remaining functions.

TRANSFORMERS ADDS MTP DECODING, TRL FIXES DEEPSEEP ZERO-3 CRASH

Transformers finally ships proper multi-token prediction support [ref:3], while TRL patches a critical DeepSpeed ZeRO-3 incompatibility that breaks Liger kernel fusion in production training.

python 46 shipped 1-min read

@huggingface 2 days ago

TRANSFORMERS CUTS DISTRIBUTED LOAD TIME IN HALF WITH SHARD-ON-READ

Transformers shipped native DTensor shard-on-read support, letting each rank load its own checkpoint slice instead of materializing the full model on every device first.

python 7 shipped 1-min read

@huggingface 3 days ago

TRL TRAINERS NOW ACCEPT DATASET DICTS DIRECTLY, KTO STREAMING GETS FASTER

TRL's stable trainers just got simpler: evaluate() now takes DatasetDict and IterableDatasetDict without forcing you to pick a split first.

python 9 shipped 1-min read

@huggingface 4 days ago

TRL FIXES MULTIMODAL LOGGING AND QUANTIZATION CRASHES ACROSS GRPO AND QORA PIPELINES

TRL shipped four critical fixes overnight for multimodal RL training, including a QLoRA + FSDP1 dtype crash that was breaking mixed-precision setups.

python 24 shipped 1-min read

— or follow the whole beat:

Elsewhere on the wire

@cachyos about 6 hours ago

CACHYOS MIRRORS HOLDING STEADY, AUR PACKAGES ROLLING UPDATES

All China mirrors are responding normally while the AUR-derived package queue processes five consecutive version bumps across developer tools.

infra 44 shipped 1-min read

Elixir & Phoenix about 6 hours ago

ELIXIR FIXES ENUM SORTER BUG THAT BROKE CUSTOM COMPARATORS

Enum.min and Enum.max were silently returning wrong results when you passed a custom sorter on ranges, and the core team just caught it.

elixir 16 shipped 1-min read

@dotnet about 6 hours ago

MATH.ROUND BUG CORRUPTS 5% OF INPUTS. EVENTPIPE LOSES HEAP DATA. SQLCLIENT GOES ASYNC

Runtime shipped a critical fix for Math.Round that was silently mangling roughly 5% of all rounding operations across the BCL, while EventPipe's data loss is finally getting a blocking mode to save your GC dumps.

dotnet 78 shipped 1-min read

@spring-projects about 6 hours ago

SPRING BOOT FIXES LAZY INITIALIZATION TRAP THAT BREAKS JPA APPS

Spring Boot patched a critical dependency loop where JPA configuration was forcing early instantiation of every AsyncTaskExecutor bean in your app, even when background bootstrapping wasn't enabled.

java 33 shipped 1-min read