Ryan dda27e50f5 [Graph Optimization] remove static_op_get_block_shape_and_split_kv_block from cudagraph (#6081) ...

* rm static_op_get_block_shape_and_split_kv_block from cudagraph

* update max_capture_shape

* fallback: zeros -> empty to avoid coverage check

* check graph_opt_config exists

* add max_capture_shape_dy2st && full_cuda_graph: false -> true in 28B vl test

* add use_cudagraph flag to control step_use_cudagraph

2026-01-20 14:05:18 +08:00

..

attention

[Graph Optimization] remove static_op_get_block_shape_and_split_kv_block from cudagraph (#6081)

2026-01-20 14:05:18 +08:00

backends

[XPU] support plugin model (#6092)

2026-01-20 13:00:09 +08:00

batch_invariant_ops

[Deterministic] Move paddle version batch invariant pkg to Fastdeploy (#4763)

2025-12-01 11:25:48 +08:00

moe

[Feature] Unify fp8 block_wise quant ops (#5991)

2026-01-15 05:50:37 -08:00

pool

[Feature] support reward model (#5301)

2025-12-02 14:55:31 +08:00

quantization

[Feature] Unify fp8 block_wise quant ops (#5991)

2026-01-15 05:50:37 -08:00

sample

[Feature]Support tag phase token enforce generation (#6034)

2026-01-15 03:59:55 -08:00

__init__.py

[LLM] First commit the llm deployment code

2025-06-09 19:20:15 +08:00

activation.py

[Feature] Unify fp8 block_wise quant ops (#5991)

2026-01-15 05:50:37 -08:00

embeddings.py

[Model] tp+ep support v1_loader (#5465)

2025-12-18 14:31:54 +08:00

linear.py

Support setting communication groups in custom_allreduce and the all-to-all\transpose fused operator during the decoding phase. (#5917)

2026-01-12 14:09:39 +08:00

lm_head.py

[Model] tp+ep support v1_loader (#5465)

2025-12-18 14:31:54 +08:00

mtp_linear.py

[Model] tp+ep support v1_loader (#5465)

2025-12-18 14:31:54 +08:00

normalization.py

fix opt qknorm (#6080)

2026-01-19 12:07:20 +08:00

pooler.py

[Feature] support reward model (#5301)

2025-12-02 14:55:31 +08:00

rotary_embedding.py

[Models] Add Qwen3-VL Model Support (#5763)

2025-12-29 17:39:33 +08:00

utils.py

[Feature] Unify fp8 block_wise quant ops (#5991)

2026-01-15 05:50:37 -08:00