main
Move test model folders (#17034)
Fix GPT2 attention scaling ignored in SDPA/FlashAttention (#44397)
[V5] Return a BatchEncoding dict from apply_chat_template by default again (#42567)