intel/auto-round

Pull Requests Commits

[pre-commit.ci] auto fixes from pre-commit.com hooks

pre-commit-ci[bot] committed 8 hours ago

b02ee329

Merge branch 'main' into copilot/add-xpu-moe-decode-implementation

a32543254 committed 8 hours ago

Verified 36562fc3

[ARK] Add varlen support for sdpa and update tests (#1957)

luoyu-intel committed 12 hours ago

Verified 948c7783

Adapt the latest sglang integration test Engine kwargs. (#1964)

changwangss committed 1 day ago

Verified 49244549

fix auto_scheme gguf format wrong bits if use tiny model (#1953)

n1ck-guo committed 1 day ago

Verified b1a6baf9

speed up awq ci (#1954)

WeiweiZhang1 committed 4 days ago

Verified 9468e52c

add other non-integer scheme support for awq (#1892)

WeiweiZhang1 committed 4 days ago

Verified 5a3ade8c

Remove kernels version limit in test (#1945)

chensuyue committed 5 days ago

Verified 142e91b4

Use shared tokenizer save helper in exporters (#1951)

changwangss committed 5 days ago

Verified 436b9db6

update ARK README (#1922)

luoyu-intel committed 5 days ago

Verified f42a0421

test: revert qwen moe to Qwen1.5-MoE-A2.7B and drop deepseek entry

Copilot committed 5 days ago

Verified 86d6e7b4

test: use Qwen3-30B-A3B in moe model perf test

Copilot committed 5 days ago

Verified 1c2c9c99

chore: remove TFLOPS column from moe decode perf test

Copilot committed 5 days ago

Verified d4dc406e

test(ark): drop base+deq(ms) column from moe decode perf

Copilot committed 5 days ago

Verified 64bb1a24

test(ark): add TFLOPS column to moe decode perf table

Copilot committed 5 days ago

Verified d7ccdf51

test(ark): add MiniMax-M2 batch=32 decode perf shapes

Copilot committed 5 days ago

Verified 68a0039d

[pre-commit.ci] auto fixes from pre-commit.com hooks

pre-commit-ci[bot] committed 5 days ago

9bda982a

Merge branch 'main' into copilot/add-xpu-moe-decode-implementation

a32543254 committed 5 days ago

Verified ac590a85

test(ark): add base+deq(ms) column to MoE perf benchmarks

Copilot committed 6 days ago

Verified 09c0a72a

test(ark): stabilize MoE prefill perf benchmark; remove misleading 100x speedups

Copilot committed 6 days ago

Verified 7094c76e

test(ark): drop "real sort" rows, keep only as-measured "real"

Copilot committed 6 days ago

Verified 70ab28ea

test(ark): replace skew/norm with empirical 256-expert distribution

Copilot committed 6 days ago

Verified 8a547f62

test(ark): replace 1K MoE prefill shapes with 8K

Copilot committed 6 days ago

Verified 72d2508a

test(ark): switch MoE prefill baseline to single torch.bmm

Copilot committed 6 days ago

Verified 8416f9f3

test(ark): add Gaussian-distribution prefill rows alongside even/skew

Copilot committed 6 days ago

Verified 045b26b3

test(ark): make even-distribution prefill rows sum exactly to seq_len*top_k

Copilot committed 6 days ago

Verified c0f2278a

test(ark): widen shape column in prefill perf table

Copilot committed 6 days ago

Verified 5f3463b0

test(ark): use MiniMax-M2 MoE config for perf shapes

Copilot committed 6 days ago

Verified 71bb9c43

test(ark): switch MoE perf shapes from Mixtral to MiniMax

Copilot committed 6 days ago

Verified e577f9bc

test(ark): set MoE prefill perf token totals to 2K and 4K

Copilot committed 6 days ago

Verified cd4f4417

Older