Default Branch

b2cb4512c5 · Create parameters overview (#1269) · Updated 2026-02-20 07:20:56 +01:00

Branches

8735931413 · Minor · Updated 2025-11-04 06:46:40 +01:00    git

4211
3948

57af463614 · Fused fused_rms+fused_rms+rope+rope (without -mqkv) · Updated 2025-11-03 17:31:56 +01:00    git

4211
3959

a7b427ffc0 · Option to enable CUDA LTO · Updated 2025-11-03 09:34:48 +01:00    git

4211
3946

3b9ace5c65 · cuda: add missing backwards RoPE op · Updated 2025-11-03 06:37:01 +01:00    git

4211
3945

b58e81d48c · Remove commented out code · Updated 2025-10-31 13:30:27 +01:00    git

4211
3946

bb4752d019 · Biased mmvq: minor optimization · Updated 2025-10-30 09:51:50 +01:00    git

4211
3941

68e7698ae8 · cohere2 - simplify graph building · Updated 2025-10-30 07:18:42 +01:00    git

4211
3951

e2b7da9684 · Minor · Updated 2025-10-28 14:24:08 +01:00    git

4211
3938

1f14f50dfd · Try removing copy indirection · Updated 2025-10-27 10:39:18 +01:00    git

4211
3935

444782523d · Make sure the bias really is 1 row to use fusion · Updated 2025-10-27 06:10:03 +01:00    git

4211
3934

a5b16b82bb · More gemv+add fusing · Updated 2025-10-26 08:32:02 +01:00    git

4211
3944

d0861f83f1 · Fix TG fused up*nary(gate) when down cannot be fused · Updated 2025-10-26 07:38:25 +01:00    git

4211
3942

6d05977940 · Change flash attention to be on by default · Updated 2025-10-25 08:32:01 +02:00    git

4211
3930

9e7d5ea64a · Diagnostics · Updated 2025-10-25 08:04:36 +02:00    git

4211
3934

17dedc8cba · Also iqk quants · Updated 2025-10-24 17:55:44 +02:00    git

4211
3935

1f96fc97c6 · Faster tensor name formatting · Updated 2025-10-24 06:42:04 +02:00    git

4211
3929

2673f55808 · fused mul+multi_add: command line argument to disable it · Updated 2025-10-23 10:34:24 +02:00    git

4211
3928

637d1b014c · Fix experts mul node name · Updated 2025-10-23 08:44:44 +02:00    git

4211
3925

3174233a9b · Various: · Updated 2025-10-22 12:02:10 +02:00    git

4211
3927

5e85b0ea51 · Also this one · Updated 2025-10-21 18:06:04 +02:00    git

4211
3924