Aleksandr Malyshev 53a30845be

Llamas 3.1 405B fp4 changes upstreaming from 355_wip (#25135) ...

Signed-off-by: Aleksandr Malyshev <maleksan@amd.com>
Co-authored-by: Aleksandr Malyshev <maleksan@amd.com>
Co-authored-by: Doug Lehr <douglehr@amd.com>

2025-09-25 19:16:53 -06:00

..

__init__.py

[feat] Support MRoPE + YaRN (#25384)

2025-09-23 03:04:47 +00:00

base.py

Llamas 3.1 405B fp4 changes upstreaming from 355_wip (#25135)

2025-09-25 19:16:53 -06:00

common.py

Remove redundant mutates_args and dispatch_key for direct_register_custom_op (#25512)

2025-09-23 22:48:40 +00:00

deepseek_scaling_rope.py

[Perf] Use FlashInfer RoPE for RotaryEmbedding.forward_cuda when available (#21126)

2025-09-19 14:06:49 -06:00

dual_chunk_rope.py

[Bugfix] Fix platform-specific routing in CustomOp implementations (#24444)

2025-09-11 17:15:01 +00:00

dynamic_ntk_alpha_rope.py

[FEAT] Refactor ROPE into module (#22192)

2025-08-04 19:12:16 -07:00

dynamic_ntk_scaling_rope.py

[FEAT] Refactor ROPE into module (#22192)

2025-08-04 19:12:16 -07:00

ernie45_vl_rope.py

[Bugfix] Fix platform-specific routing in CustomOp implementations (#24444)

2025-09-11 17:15:01 +00:00

linear_scaling_rope.py

[FEAT] Refactor ROPE into module (#22192)

2025-08-04 19:12:16 -07:00

llama3_rope.py

[FEAT] Refactor ROPE into module (#22192)

2025-08-04 19:12:16 -07:00

llama4_vision_rope.py

[Perf] Use FlashInfer RoPE for RotaryEmbedding.forward_cuda when available (#21126)

2025-09-19 14:06:49 -06:00

mrope.py

[feat] Support MRoPE + YaRN (#25384)

2025-09-23 03:04:47 +00:00

ntk_scaling_rope.py

[FEAT] Refactor ROPE into module (#22192)

2025-08-04 19:12:16 -07:00

phi3_long_rope_scaled_rope.py

[FEAT] Refactor ROPE into module (#22192)

2025-08-04 19:12:16 -07:00

rocm_aiter_rope_ops.py

Llamas 3.1 405B fp4 changes upstreaming from 355_wip (#25135)

2025-09-25 19:16:53 -06:00

yarn_scaling_rope.py

[FEAT] Refactor ROPE into module (#22192)

2025-08-04 19:12:16 -07:00