GPT2Model model output inconsistency between different transformers versions

John6666 · March 22, 2025, 4:55am

Perhaps this? SDPA is now default attention.

[`GPT2`] Add SDPA support (#31172)

committed 07:40AM - 19 Jun 24 UTC

* `gpt2` sdpa support * fix (at least) one test, style, repo consistency *… fix sdpa mask in forward --> fixes generation * test * test2 * test3 * test4 * simplify shapes for attn mask creation and small comments * hub fail test * benchmarks * flash attn 2 mask should not be inverted on enc-dec setup * fix comment * apply some suggestion from code review - only save _attn_implentation once - remove unnecessary comment * change elif logic * [run-slow] gpt2 * modify `test_gpt2_sample_max_time` to follow previous assertion patterns

Topic		Replies	Views
Inconsistent GPT2Model results between transformers versions Intermediate	7	30	July 19, 2025
Difference in output logits when using a subsection of the input sentence 🤗Transformers	0	372	January 8, 2023
Different model.generate() predictions between batched and unbatched/padded token inputs 🤗Transformers	2	2265	August 26, 2023
Hidden States of OpenAI GPT2 inconsistent 🤗Transformers	2	284	October 25, 2021
Looks like the new transformer 4.49.0 has some issues 🤗Transformers	3	278	March 6, 2025

GPT2Model model output inconsistency between different transformers versions

Related topics