huggingface/transformers

Pull Requests Commits

Merge branch 'main' into cb-training-turn

qgallouedec committed 53 days ago

Verified 2804f33a

Let a thread that shares the model take a turn on it under continuous batching

qgallouedec committed 53 days ago

3269558f

[LongcatFlash] Fix test_longcat_generation_cpu: use device_map="cpu" to avoid MoE disk offload issue (#48377)

ydshieh committed 53 days ago

Verified 83d024e1

[Qwen3VLMoe] Update `test_small_model_integration_test_batch` expected output (value drift) (#48376)

ydshieh committed 53 days ago

Verified 155b8993

[ONNX] Skip affected models on torch 2.13 (two dynamo regressions) (#48191)

ydshieh committed 53 days ago

Verified 19876312

Fix `safe_open` mmap memory exhaustion on Windows by using `pread` backend (#48341)

eryk-roch committed 53 days ago

Verified 8631167e

Retry get_daily_ci_runs on stale GitHub API cache (#48374)

ydshieh committed 53 days ago

Verified aa372615

Fix Zamba2 construction for num_mem_blocks > 1 checkpoints (#48325)

john-rocky committed 53 days ago

Verified 323f2434

Implement VibeVoice (#40546)

pengzhiliang committed 53 days ago

Verified 640a08a5

[Docs] Change 5.3 Flash pos in toc (#48366)

Dovis01 committed 53 days ago

Verified afa80816

[conftest] Use get_cpu_ram_total_gib for psutil patch (cgroup-aware) (#48290)

ydshieh committed 53 days ago

Verified 35a66fa6

Fix flaky test_training_gradient_checkpointing for BigBirdPegasus (fp noise filter) (#48332)

ydshieh committed 53 days ago

Verified 9b4f0a0c

Quiet continuous batching at default verbosity (#48314)

qgallouedec committed 53 days ago

Verified 2e7033ed

Wait for the first request in the async continuous batching bootstrap (#48304)

qgallouedec committed 53 days ago

Verified 01846b52

[CB] Fail faster (#48334)

remi-or committed 54 days ago

Verified b6c0bfe0

Ignore a stale best checkpoint recorded in a resumed trainer state (#48319)

VaggelisGian committed 54 days ago

Verified 36deb0b5

Keep MXFP4 weights quantized on XPU when use_kernels is set (#47923)

jiqing-feng committed 54 days ago

Verified e15ea9b9

Fix missing FP8 TP layer overrides (#48343)

changwangss committed 54 days ago

Verified d65f37cc

[docs] Fix links and remove TokenizerFast (#47748)

stevhliu committed 54 days ago

Verified 762fae6e

Fix incorrect token classification prefix for ESMC (#48348)

Rocketknight1 committed 54 days ago

Verified 4e5538a3

Create the continuous batching CPU group with local synchronization (#48302)

qgallouedec committed 54 days ago

Verified dabae5fc

Resolve continuous batching config against the text config for composite models (#48299)

qgallouedec committed 54 days ago

Verified 598d8ba8

[Glm 5.3 Flash] GLM 5.3 Flash Support (#48342)

Dovis01 committed 54 days ago

Verified eb4d9e2a

Fix kernel commit and repo paths for ESMFold2 (#48186)

Rocketknight1 committed 54 days ago

Verified ba5cb08d

[`CI`] Unblock fast CI for now (failing tests) (#48344)

vasqu committed 54 days ago

Verified 1e452995

[qwen4_exp] disable torch/onnx export tests due to data-dependent control flow (#48345)

ydshieh committed 54 days ago

Verified 0f0f7e1a

[debug] Trace previous CI run selection in get_previous_daily_ci.py (#48338)

ydshieh committed 54 days ago

Verified 36bc98ef

Normalize HunYuanVL's legacy field aliases via attribute_map (#48261)

hmellor committed 54 days ago

Verified 2c889bf2

Add Qwen4Exp model (#48337)

Cyrilvallez committed 54 days ago

Verified fc5c5bde

Restore BC for the tensor-parallel API (#48300)

ArthurZucker committed 54 days ago

Verified f0e42efb

Older