Woosuk Kwon 90d43db442 [Optimization] Truncate kv page indices for sliding window attention ...

Signed-off-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>

2025-08-15 19:10:01 +00:00

..

mla

[Core] Allow full cudagraph with separate attention routines and orthogonal to compilation, add support for FA2 and FlashInfer (#20059)

2025-08-15 10:01:39 -04:00

__init__.py

[V1] Implement vLLM V1 [1/N] (#9289)

2024-10-22 01:24:07 -07:00

cpu_attn.py

[v1][attention] Support Hybrid Allocator + FlashInfer (#21412)

2025-07-30 01:45:29 +00:00

flash_attn.py

[Core] Allow full cudagraph with separate attention routines and orthogonal to compilation, add support for FA2 and FlashInfer (#20059)

2025-08-15 10:01:39 -04:00

flashinfer.py

[Optimization] Truncate kv page indices for sliding window attention

2025-08-15 19:10:01 +00:00

flex_attention.py

Support encoder_only attention for FlexAttention (#22273)

2025-08-06 18:37:14 -07:00

linear_attn.py

[V1] [Hybrid] Support Minimax-Text-01 in V1 (#22151)

2025-08-08 23:08:48 -07:00

mamba1_attn.py

[V1] - Split Prefill and Decode for Mamba1 models (#22653)

2025-08-15 08:59:52 +00:00

mamba2_attn.py

[Core] Allow full cudagraph with separate attention routines and orthogonal to compilation, add support for FA2 and FlashInfer (#20059)

2025-08-15 10:01:39 -04:00

mamba_selectors.py

[Mamba] - refactor: Renamed mamba_attn to mamba2_attn (#22818)

2025-08-15 06:38:05 +00:00

pallas.py

[Attention] Clean up iRoPE in V1 (#21188)

2025-07-21 09:10:30 -07:00

rocm_aiter_fa.py

[Core] Allow full cudagraph with separate attention routines and orthogonal to compilation, add support for FA2 and FlashInfer (#20059)

2025-08-15 10:01:39 -04:00

tree_attn.py

[V1] Add tree drafting tests for eagle spec decoding (#22705)

2025-08-13 04:11:28 -07:00

triton_attn.py

[Core] Allow full cudagraph with separate attention routines and orthogonal to compilation, add support for FA2 and FlashInfer (#20059)

2025-08-15 10:01:39 -04:00

utils.py

[Core] Allow full cudagraph with separate attention routines and orthogonal to compilation, add support for FA2 and FlashInfer (#20059)

2025-08-15 10:01:39 -04:00

xformers.py

[V1] port xformers backend to v1 (#21342)

2025-08-05 10:04:46 -07:00