transformers

2ef59646 - Fix `max_length_q` and `max_length_k` types to `flash_attn_varlen_func` (#37206)

Fix `max_length_q` and `max_length_k` types to `flash_attn_varlen_func` (#37206) Also add notes asking users to set `TORCHDYNAMO_CAPTURE_SCALAR_OUTPUTS=1` or call `torch._dynamo.config.capture_scalar_outputs = True`, as currently this will cause a graph break. Signed-off-by: Hollow Man <hollowman@opensuse.org>