All issues

Search results

[serge] integration failure triage - 2026-08-28

Status: Open.

#48403 In huggingface/transformers;

· github-actions opened on Aug 28, 2026

Nested per-layer rope_parameters is misparsed when layer_types omits one of its keys

Status: Open.

#48392 In huggingface/transformers;

· albertvillanova opened on Aug 28, 2026

Vectorized (batched) logits processors for assisted decoding

Status: Open.

#48390 In huggingface/transformers;

· VaggelisGian opened on Aug 28, 2026

More freedom to override kernels for functions

bug

Status: Open.

#48375 In huggingface/transformers;

· zucchini-nlp opened on Aug 27, 2026

NVFP4 MoE experts are dequantized without weight_global_scale, silently

Status: Open.

#48371 In huggingface/transformers;

· Bluear7878 opened on Aug 27, 2026

适配鸿蒙

Feature requestRequest for a new featureRequest for a new feature

Status: Open.

#48369 In huggingface/transformers;

· Hao-bigMan opened on Aug 27, 2026

FineGrainedFP8HfQuantizer.update_tp_plan crashes when the model has no tp_plan

Status: Open.

#48351 In huggingface/transformers;

· 01554 opened on Aug 26, 2026

qwen4_exp: fp8-quantized n-gram (PLE) embedding rows are gathered without dequantization

Status: Open.

#48350 In huggingface/transformers;

· 01554 opened on Aug 26, 2026

FineGrainedFP8: modules_to_not_convert never matches on text-only loads of multimodal checkpoints (silent quality corruption)

Status: Open.

#48349 In huggingface/transformers;

· 01554 opened on Aug 26, 2026

QA: mtp models are gone from hub

bug

Status: Open.

#48328 In huggingface/transformers;

· tarekziade opened on Aug 26, 2026

Chunk conversion op silently produces unequal chunks when the fused tensor is not divisible by the target count

Status: Open.

#48323 In huggingface/transformers;

· VaggelisGian opened on Aug 25, 2026

Continuous batching CUDA-graph capture crashes under distributed runs (allocator internal assert)

Status: Open.

Bug

#48312 In huggingface/transformers;

· qgallouedec opened on Aug 25, 2026