transformers
418 issues found
incompatibility with torch 2.4.1
6/9/2026
[Continuous batching] Tracker for issues and feature requests
6/9/2026
Investigate training support for Sapiens2ForPoseEstimation when labels are provided
6/9/2026
StopStringCriteria misses CJK stop strings on byte-level tokenizers when a character splits into byte-fragment tokens
6/9/2026
DeepSeek-Coder v1 tokenizer produces incorrect output on transformers v5+ (gap in PR #44801's fix)
6/8/2026
CodeLlama tokenizer strips one leading space on encode→decode round-trip (regression vs v4)
6/8/2026
difference in performance of sam3 video model compared to original sam3 implementation
6/8/2026
Different results for PPDocLayoutV3 on CPU and CUDA
6/8/2026
Your docs show how to use models but not what they cost
6/8/2026
from_config(dtype=...) builds weights in the requested dtype but leaves model.config.dtype stale
6/8/2026
[Bug] Official Nemotron example broken
6/7/2026
New RAG framework: VORTEXRAG — 7-layer causal RAG achieving EM 74.8, Faithfulness 0.94
6/7/2026
Suggestion: Give a warning for 4D attention mask which use float dtype but in 0.0 and 1.0 term (instead of -inf and zero).
6/6/2026
`dtype` is silently ignored when loading composite (multimodal) checkpoints through AutoModelForCausalLM (affects all Qwen3.5 repos)
6/6/2026
[Bug] Incorrect error message for T5 model family's decoder input validation
6/6/2026
ValueError: Cannot use apply_chat_template because this processor does not have a chat template.
6/5/2026
Static cache `early_initialization` corrupts linear-attention layers (wrong shapes → `RuntimeError`)
6/5/2026
Qwen3.5 crash with `cache_implementation="static"` — `KeyError: 'linear_attention'`
6/5/2026
Vision RoPE refactor
6/5/2026
`added_tokens_encoder` rebuilds the entire added-token map on every access
6/4/2026