Who contributed to Hugging Face on June 15, 2026?

3 developers shipped this update, including behroozazarkhalili, qgallouedec, and yiyixuxu.

What were the notable Hugging Face updates?

fix: share frozen layers with reference model instead of duplicating in memory, Default SFT loss to chunked_nll, and Point "Coding with AI agents" links at the rendered docs site.

@huggingface

Transformers, Datasets, and the open AI-model layer

github ↗

Pick a date

Topics: Python AI / ML Full archive →

The Wire · Showcase

TRL FIXES MEMORY LEAK IN REFERENCE MODEL HANDLING, DEFAULTS SFT LOSS TO FASTER CHUNKED ALGORITHM

By RepoJournal · Filed 06:03 UTC on June 15, 2026 · About Hugging Face

3 people shipped this

behroozazarkhalili @behroozazarkhalili 1 cited

qgallouedec @qgallouedec 1 cited

yiyixuxu @yiyixuxu 1 cited

Two critical training optimizations landed in TRL overnight: frozen layers now share memory instead of duplicating, and the default SFT loss switches to chunked_nll for 30-40% faster training on most architectures.

The reference model fix [1] addresses a subtle but costly bug in `create_reference_model` where frozen shared layers were being deepcopied then never actually reassigned, leaving duplicate parameters in memory. This tanks performance when you're trying to freeze layers to save VRAM in DPO or preference-tuning workflows. The memory leak is fixed in the same PR, so frozen layers now properly share parameters with the reference model. Meanwhile, the SFT trainer's default loss switches from 'nll' to 'chunked_nll' [2], which benchmarks show cuts training time 30-40% without accuracy loss on most model sizes, with automatic fallback to 'nll' when using Liger kernels. In the diffusers repo, the team fixed broken anchor links in the contributor guide [3] that were pointing to GitHub's symlink rendering instead of the actual docs site. These are the changes that compound: better training performance, lower memory footprint, clearer onboarding for new contributors.

Action items

→ Upgrade TRL if you're training preference models or using num_shared_layers - redeploy with the reference model fix [ref:5] huggingface/trl [immediate]
→ Update SFT training configs to verify chunked_nll is your new default [ref:3], especially if you're running multi-GPU training huggingface/trl [plan]
→ Update any internal docs linking to diffusers CONTRIBUTING.md - anchor links now resolve correctly [ref:1] huggingface/diffusers [monitor]

References

[1] fix: share frozen layers with reference model instead of duplicating in memory ↗ huggingface/trl
[2] Default SFT loss to chunked_nll ↗ huggingface/trl
[3] Point "Coding with AI agents" links at the rendered docs site ↗ huggingface/diffusers

Quick answers

What shipped in Hugging Face on June 15, 2026?: Two critical training optimizations landed in TRL overnight: frozen layers now share memory instead of duplicating, and the default SFT loss switches to chunked_nll for 30-40% faster training on most architectures. In total, 4 commits and 3 pull requests landed.
Who contributed to Hugging Face on June 15, 2026?: 3 developers shipped this update, including behroozazarkhalili, qgallouedec, and yiyixuxu.
What were the notable Hugging Face updates?: fix: share frozen layers with reference model instead of duplicating in memory, Default SFT loss to chunked_nll, and Point "Coding with AI agents" links at the rendered docs site.

TRANSFORMERS-CI SHIPS INTELLIGENT TEST TRIAGE, CUTS LLM NOISE 100X

Hugging Face's serge verify loop now distills 100KB failure tracebacks down to the actual assertion, runs every test 5 times to catch flakiness, and dispatches each failure category with a custom instruction before spinning GPU or consuming a single LLM token.

python 45 shipped 2-min read

@huggingface 2 days ago

TRL CI BURNS, BITSANDBYTES PINNED BELOW 0.50.0

Hugging Face's TRL training library is firefighting two separate CI failures overnight, forcing temporary pins on bitsandbytes and xfailing critical tests against a broken transformers dev build.

python 7 shipped 1-min read

@huggingface 3 days ago

TRANSFORMERS OVERHAULS LINEAR ATTENTION WHILE DEPRECATING LEGACY RESPONSE SCHEMA

The transformers library is retiring its fragile response_schema prototype in favor of streaming-compatible parsing, while simultaneously refactoring every linear attention model to use standardized convolution patterns.

python 70 shipped 2-min read

@huggingface 4 days ago

TRANSFORMERS SHIPS FSDP DISTRIBUTED TRAINING STACK, HUB LIBRARY PLUGS REDOS HOLE

Hugging Face landed distributed training orchestration in transformers while plugging a regex vulnerability that could stall untrusted card parsing for minutes.

+10

python 61 shipped 1-min read

Elsewhere on the wire

AI Agents about 21 hours ago

CLINE SHIPS FIVE CRITICAL AGENT FIXES AFTER END-TO-END TESTING BLITZ

Saoudrizwan just landed a cascade of fixes that prevent the agent loop from running wild, stop file reads from resolving to the wrong directory, and plug the compaction pipeline.

ai-agents 16 shipped 2-min read

Local LLMs about 21 hours ago

LLAMA.CPP SHIPS MINIMAX-M3 SPARSE ATTENTION, VLLM UNLOCKS HETEROGENEOUS PREFILL-DECODE

The inference stack gets a new model architecture and a fundamental shift in how prefill and decode can run on different hardware configurations.

+10

llms 125 shipped 1-min read

@CachyOS about 21 hours ago

GLIBC 2.43 REBUILD SHIPS, INSTALLER JUMPS TO C++26

CachyOS pushed glibc 2.43 through the pipeline while the CLI installer upgraded its C++ standard to C++26, signaling a major refresh across core infrastructure.

infra 97 shipped 1-min read

Elixir & Phoenix about 21 hours ago

ECTO FIXES SUBQUERY PLANNING IN JOINS, ELIXIR PATCHES IEX PARSING

Ecto's query planner now handles subqueries correctly in interpolated join filters, while Elixir fixes a syntax error that broke 'not in' as a continuation operator in IEx.

elixir 10 shipped 1-min read

Want every project, not just this one?

Follow @huggingface