feat: add top_k_top_p_sampling_from_probs_v248320 definition and workloads (synced from upstream)

feat: add Qwen3.5 model tag for gdn_decode_qk8_v16_d128_k_last (#204)

llama1b (#303)

feat(kimi-k2.5): sampling defs (v163840) + MLA tag updates (#310)

Sync definitions and tests from flashinfer-bench flashinfer_trace (#308)

Add top_k_sampling_from_probs_v202048: baseline solution + workloads (#266)

Add top_p_sampling_from_probs_v202048: baseline solution + workloads (#267)

Add top_k_top_p_sampling_from_probs_v202048: baseline solution + workloads (#268)

Add gqa_paged_prefill_causal_h24_kv8_d128_ps1: solution + workloads + blobs + def + tests (#250)

Add gqa_ragged_prefill_causal_h24_kv8_d128: solution + workloads + blobs + def + tests (#254)

Add gqa_paged_prefill_causal_h24_kv8_d128_ps64: solution + workloads + blobs + def + tests (#253)

Add gqa_paged decode/prefill workloads for Llama 4 Scout 17B-16E (TP=8) (#259)

Vectorize dsa_sparse_attention_h16_ckv512_kpe64_topk2048_ps64 reference (#258)

Add gqa_paged_decode_h24_kv8_d128_ps1: solution + workloads + blobs + eval trace (Llama 3.2 3B) (#244)

Add gqa_paged_prefill_causal_h16_kv2_d128_ps1: workloads + baseline solution + eval trace (#235)

Add gqa_ragged_prefill_causal_h8_kv1_d256: definition + solution + 20 workloads + blobs + traces (Qwen3-Next-80B TP=2) (#239)

Add gqa_paged_prefill_causal_h16_kv2_d128_ps64: workloads + baseline solution + eval trace (#236)

Add gqa_paged_decode_h16_kv2_d128_ps64: workloads + baseline solution + eval trace (#234)

Add gqa_ragged_prefill_causal_h16_kv2_d128 (Qwen3-32B TP=4 ragged prefill) (#238)

Add gqa_paged_decode_h16_kv2_d128_ps1: workloads + baseline solution + eval trace (#233)

Add gqa_paged_decode_h32_kv4_d128_ps64: solution + workloads + blobs + eval trace (Qwen3-30B-A3B) (#243)

Add gqa_paged_prefill_causal_h32_kv4_d128_ps64: solution + workloads + blobs + eval trace (Qwen3-30B-A3B) (#242)

Add gqa_paged_prefill_causal_h40_kv10_d128_ps1: solution + workloads + blobs + eval trace (20/20 PASSED) (#229)

Add mla_ragged_prefill_causal_h16_qk192_vo128: solution + workloads + blobs + eval trace (#225)

feat: add gqa_paged_prefill_causal_h24_kv4_d128_ps64 (Mixtral 8x22B TP=2 prefill) (#196)

feat: add gqa_paged_decode_h24_kv4_d128_ps64 workloads (Mixtral 8x22B TP=2) (#195)

feat: add gqa_paged_prefill_causal_h24_kv4_d128_ps1 workloads, solution, and definition (#199)

Add gqa_paged_decode_h48_kv8_d128_ps1: solution + workloads + def + tests (#162)

sync: flashinfer_trace from flashinfer-bench origin/main (#124)

Re-collect gqa_paged h=20 workloads from SGLang (Qwen3-14B TP=2) (#37)

sycn definitions from github flashinfer-bench (#24)

gdn workload update: all tensors are dumped; add flashinfer gdn baseline (#21)

fix: gdn workload path and shape (#19)

fix: error when copying definition (#18)

fix: real k for gdn prefill (tp4) (#14)

Sync with FlashInfer-Bench (#9)

Update DSA definition topk to 2048 instead of 256

upd

add dsa/gdn/ps64 kernels

update decode axes name

Added FlashInfer Trace