apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2026-04-23 00:17:25 +08:00
Code Issues Actions 19 Packages Projects Releases Wiki Activity
Files
1fb81941910e77750c02936c6f7396061c36ec15
FastDeploy/custom_ops/gpu_ops
T
History
AIbin 1fb8194191 [OP][Models][Optimization] 优化 RoPE CUDA kernel 并更新 DeepSeek V3 配置 (#7359) 
* dsk del prefill mask

* dsk support 1M+ seq_len rope

* update rope tests

* Replace max_position_embeddings with max_model_len

* 1D grid: gridDim.x has a maximum size of 2^31-1, far exceeding the actual number of tokens.
2026-04-13 19:12:36 +08:00
..
append_attn
[RL] change glm rope_emb calculation (#7316)
2026-04-11 18:36:28 +08:00
common
init (#6642)
2026-03-04 21:55:31 +08:00
custom_all_reduce
[BugFix][Optimization] Replace silent failures with catchable exceptions and informative error messages (#6533)
2026-03-16 21:32:43 +08:00
cutlass_extensions
init (#6642)
2026-03-04 21:55:31 +08:00
cutlass_kernels
support moe for sm103 (#7238)
2026-04-08 15:52:39 +08:00
flash_mask_attn
[BugFix] Fix Async D2H copy bug & flash mash atten cache V out of bound bug (#7221)
2026-04-10 11:31:51 +08:00
fp8_gemm_with_cutlass
init (#6642)
2026-03-04 21:55:31 +08:00
glog
init (#6642)
2026-03-04 21:55:31 +08:00
int8_gemm_with_cutlass
init (#6642)
2026-03-04 21:55:31 +08:00
machete
[BugFix] detection jinja2 (#7251)
2026-04-09 11:30:16 +08:00
mla_attn
[BugFix][Optimization] Replace silent failures with catchable exceptions and informative error messages (#6533)
2026-03-16 21:32:43 +08:00
moba_attn
init (#6642)
2026-03-04 21:55:31 +08:00
moe
[OP]Unify MoE op with moe_permute path for bf16 GLM (#7164)
2026-04-09 16:17:56 +08:00
quantization
cuda13.0, implement changes to CCCL (#6751)
2026-03-10 16:47:02 +08:00
sample_kernels
cuda13.0, implement changes to CCCL (#6751)
2026-03-10 16:47:02 +08:00
sparse_indexer
fix dsa (#7252)
2026-04-08 20:21:38 +08:00
speculate_decoding
[Optimization]【Hackathon 10th Spring No.49】GPU ngram_match: BlockScan Phase 2 -optimized (#7136)
2026-04-07 01:36:25 -07:00
w4afp8_gemm
[BugFix] fix wint2 (#6109)
2026-01-20 21:46:21 +08:00
wfp8afp8_sparse_gemm
init (#6642)
2026-03-04 21:55:31 +08:00
append_attention.cu
[RL] RoPE without fmad opt (#6901)
2026-03-24 21:19:53 +08:00
beam_search_softmax.cu
cuda13.0, implement changes to CCCL (#6751)
2026-03-10 16:47:02 +08:00
cccl_compat.h
cuda13.0, implement changes to CCCL (#6751)
2026-03-10 16:47:02 +08:00
cpp_extensions.cc
[OP]Unify MoE op with moe_permute path for bf16 GLM (#7164)
2026-04-09 16:17:56 +08:00
cuda_multiprocess.h
[BugFix][Optimization] Replace silent failures with catchable exceptions and informative error messages (#6533)
2026-03-16 21:32:43 +08:00
custom_ftok.h
[BugFix] replace ftok with custom_ftok in get_output/save_output ops (#6822)
2026-03-16 14:22:18 +08:00
dequant_int8.cu
init (#6642)
2026-03-04 21:55:31 +08:00
enforce_generation.cu
init (#6642)
2026-03-04 21:55:31 +08:00
env.h
init (#6642)
2026-03-04 21:55:31 +08:00
fused_get_rotary_embedding.cu
init (#6642)
2026-03-04 21:55:31 +08:00
fused_hadamard_quant_fp8.cu
init (#6642)
2026-03-04 21:55:31 +08:00
fused_mask_swiglu_fp8_quant_kernel.cu
[Optimization] optimize fused_swiglu_fp8_quant_kernel (#7007)
2026-03-27 16:10:16 +08:00
fused_neox_rope_embedding.cu
[Feature] Optim PaddleOCR-VL (#4873)
2025-11-07 14:56:44 +08:00
fused_rotary_position_encoding.cu
[OP][Models][Optimization] 优化 RoPE CUDA kernel 并更新 DeepSeek V3 配置 (#7359)
2026-04-13 19:12:36 +08:00
gather_idx.cu
init (#6642)
2026-03-04 21:55:31 +08:00
gelu_tanh.cu
[Metax] adapt to the latest develop (#6282)
2026-01-29 23:21:20 -08:00
get_attn_mask_q.cu
[Optimization] Support FA2/FA3/FA4 with attn_mask_q (#6354)
2026-02-05 14:39:00 +08:00
get_data_ptr_ipc.cu
[BugFix][Optimization] Replace silent failures with catchable exceptions and informative error messages (#6533)
2026-03-16 21:32:43 +08:00
get_img_boundaries.cc
[Models] Add Qwen3-VL Model Support (#5763)
2025-12-29 17:39:33 +08:00
get_mm_split_fuse.cc
init (#6642)
2026-03-04 21:55:31 +08:00
get_output_ep.cc
[BugFix] replace ftok with custom_ftok in get_output/save_output ops (#6822)
2026-03-16 14:22:18 +08:00
get_output_msg_with_topk.cc
[BugFix] replace ftok with custom_ftok in get_output/save_output ops (#6822)
2026-03-16 14:22:18 +08:00
get_output.cc
[BugFix] replace ftok with custom_ftok in get_output/save_output ops (#6822)
2026-03-16 14:22:18 +08:00
get_padding_offset_system.cu
init (#6642)
2026-03-04 21:55:31 +08:00
get_padding_offset.cu
fix empty get_padding_offset (#6462)
2026-02-12 12:34:23 +08:00
get_position_ids_and_mask_encoder_batch.cu
init (#6642)
2026-03-04 21:55:31 +08:00
helper.cu
[RL] RoPE without fmad opt (#6901)
2026-03-24 21:19:53 +08:00
helper.h
[RL] RoPE without fmad opt (#6901)
2026-03-24 21:19:53 +08:00
init_signal_layerwise.cc
init (#6642)
2026-03-04 21:55:31 +08:00
ipc_sent_key_value_cache_by_remote_ptr.cu
[BugFix][Optimization] Replace silent failures with catchable exceptions and informative error messages (#6533)
2026-03-16 21:32:43 +08:00
limit_thinking_content_length.cu
[OP][Feature] 统一 limit_thinking_content_length CUDA 算子，支持回复长度限制与注入序列 (#6493)
2026-02-25 21:36:50 +08:00
merge_prefill_decode_output.cu
[Optimization] [OP] [Models] dsk del prefill mask (#7313)
2026-04-11 19:32:27 +08:00
msg_utils.h
init (#6642)
2026-03-04 21:55:31 +08:00
multi_head_latent_attention.cu
【Hackathon 9th No.86】autogen MoeFastHardamardImplWrapper template_instantiation (#4592)
2025-10-30 10:28:36 +08:00
ngram_mask.cu
init (#6642)
2026-03-04 21:55:31 +08:00
noaux_tc_redundant.cu
[BugFix]fix handle 4 return values from noaux_tc_redundant op (#6384)
2026-02-09 13:17:47 +08:00
noaux_tc.cu
init (#6642)
2026-03-04 21:55:31 +08:00
noauxtc_kernel.h
[Others] clean code (#5691)
2025-12-24 11:28:47 +08:00
open_shm_and_get_meta_signal.cc
init (#6642)
2026-03-04 21:55:31 +08:00
per_token_quant_fp8.cu
[Feature] Support EP prefill with num_worst_tokens (#6574)
2026-03-11 17:09:07 +08:00
read_data_ipc.cu
[BugFix][Optimization] Replace silent failures with catchable exceptions and informative error messages (#6533)
2026-03-16 21:32:43 +08:00
read_ids.py
polish code with new pre-commit rule (#2923)
2025-07-19 23:19:27 +08:00
read_temp_ids.py
polish code with new pre-commit rule (#2923)
2025-07-19 23:19:27 +08:00
reasoning_phase_token_constraint.cu
[Feature] Support mtp overlap schedule (#7001)
2026-04-01 14:24:26 +08:00
rebuild_padding.cu
[Feature] Support mtp overlap schedule (#7001)
2026-04-01 14:24:26 +08:00
recover_decode_task.cu
init (#6642)
2026-03-04 21:55:31 +08:00
remote_cache_kv_ipc.cc
init (#6642)
2026-03-04 21:55:31 +08:00
remote_cache_kv_ipc.h
[BugFix] replace ftok with custom_ftok in get_output/save_output ops (#6822)
2026-03-16 14:22:18 +08:00
save_output_msg_with_topk.cc
[BugFix] replace ftok with custom_ftok in get_output/save_output ops (#6822)
2026-03-16 14:22:18 +08:00
save_with_output_msg.cc
[BugFix] replace ftok with custom_ftok in get_output/save_output ops (#6822)
2026-03-16 14:22:18 +08:00
save_with_output_msg.h
init (#6642)
2026-03-04 21:55:31 +08:00
save_with_output.cc
init (#6642)
2026-03-04 21:55:31 +08:00
scaled_gemm_f8_i4_f16_gemm.cu
init (#6642)
2026-03-04 21:55:31 +08:00
scaled_gemm_f8_i4_f16_weight_quantize.cu
init (#6642)
2026-03-04 21:55:31 +08:00
seqs2seqs.cu
init (#6642)
2026-03-04 21:55:31 +08:00
set_data_ipc.cu
[BugFix][Optimization] Replace silent failures with catchable exceptions and informative error messages (#6533)
2026-03-16 21:32:43 +08:00
set_flags.cu
init (#6642)
2026-03-04 21:55:31 +08:00
set_mask_value.cu
init (#6642)
2026-03-04 21:55:31 +08:00
set_value_by_flags_and_idx.cu
[Feature] GPU Memory Optimization and Retirement of V0 Scheduler (#6407)
2026-02-28 15:07:43 +08:00
share_external_data.cu
[BugFix][Optimization] Replace silent failures with catchable exceptions and informative error messages (#6533)
2026-03-16 21:32:43 +08:00
step_reschedule.cu
[BugFix] replace ftok with custom_ftok in get_output/save_output ops (#6822)
2026-03-16 14:22:18 +08:00
step_system_cache.cu
init (#6642)
2026-03-04 21:55:31 +08:00
step.cu
cuda13.0, implement changes to CCCL (#6751)
2026-03-10 16:47:02 +08:00
stop_generation_multi_ends.cu
[Feature] GPU Memory Optimization and Retirement of V0 Scheduler (#6407)
2026-02-28 15:07:43 +08:00
stop_generation.cu
init (#6642)
2026-03-04 21:55:31 +08:00
swap_cache_batch.cu
[BugFix][Optimization] Replace silent failures with catchable exceptions and informative error messages (#6533)
2026-03-16 21:32:43 +08:00
swap_cache_layout.cu
[KVCache] Storage cache supports c8 model (#6298)
2026-02-06 12:01:17 +08:00
swap_cache.cu
[BugFix][Optimization] Replace silent failures with catchable exceptions and informative error messages (#6533)
2026-03-16 21:32:43 +08:00
system2group.cu
init (#6642)
2026-03-04 21:55:31 +08:00
text_image_gather_scatter.cu
[Graph Optimization] Fix IR graph dependency error exposed after enabling SOT by updating the return value of TextImageGatherScatter (#4610)
2025-10-28 18:31:23 +08:00
text_image_index_out.cu
init (#6642)
2026-03-04 21:55:31 +08:00
token_penalty_multi_scores.cu
[Feature][BugFix][OP] Enhance Deterministic Inference Mode with Kernel-level Fixes and Batch-invariant BMM (#6610)
2026-03-09 10:27:53 +08:00
token_penalty_only_once.cu
init (#6642)
2026-03-04 21:55:31 +08:00
token_transfer.hpp
init (#6642)
2026-03-04 21:55:31 +08:00
transfer_output.cc
init (#6642)
2026-03-04 21:55:31 +08:00
tune_cublaslt_gemm.cu
[BugFix][Optimization] Replace silent failures with catchable exceptions and informative error messages (#6533)
2026-03-16 21:32:43 +08:00
unset_data_ipc.cu
init (#6642)
2026-03-04 21:55:31 +08:00
update_attn_mask_offsets.cu
[Speculative Decoding] Optimize attn_mask_offset and fix mtp bug (#7005)
2026-03-25 01:52:06 -07:00
update_inputs_beam.cu
[XPU] bind some OPs for VL model with pybind (#4522)
2025-10-27 10:50:08 +08:00
update_inputs_v1.cu
[Model Runner] Refactor execute_model for GPU async scheduling (#6176)
2026-01-28 14:19:33 +08:00
update_inputs.cu
[Model Runner] Refactor execute_model for GPU async scheduling (#6176)
2026-01-28 14:19:33 +08:00
update_split_fuse_input.cu
init (#6642)
2026-03-04 21:55:31 +08:00