Enable SDPA for Decision Transformer (#49550)
* Enable SDPA / Flash Attention flags for Decision Transformer
DecisionTransformerGPT2Attention already dispatches through ALL_ATTENTION_FUNCTIONS (copied from GPT2), but the PreTrainedModel classes did not declare support, so attn_implementation="sdpa" raised a ValueError. Declare the same flags as GPT2PreTrainedModel and pass the extra model inputs to the common SDPA equivalence tests.
Assisted-by: Claude:claude-opus-5-5
* Drop the Flash Attention flag and a broken test comment
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>