biondizzle
  • Joined on 2025-12-10
biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-24 22:19:17 +00:00
18f3274c0b D1: DEBUG - NO-OP O rescale (multiply by 1.0) to test TMEM round-trip
biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-24 22:18:14 +00:00
c33185ca0a D1: add rescale diagnostic
biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-24 22:17:25 +00:00
02edff5ac7 D1: add KV merge test using log-sum-exp (avoids TMEM round-trip)
biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-24 22:15:40 +00:00
0f30319e06 Revert "D1: move O rescale atoms outside const_expr guard (match CUTLASS pattern)"
biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-24 22:07:20 +00:00
aaf21d8ac1 D1: move O rescale atoms outside const_expr guard (match CUTLASS pattern)
biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-24 22:04:53 +00:00
35a3c04e8e fix debug test
biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-24 22:04:22 +00:00
a391aa1fd3 D1: add rescale debug test
biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-24 22:02:57 +00:00
55c6903980 D1: fix O rescale identity tensor - use PV MMA shape not QK shape
biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-24 22:00:50 +00:00
f1aab1bfc1 D1: add multi-KV-tile O rescale test (s_k=256,384,512)
biondizzle deleted tag v-tma-multitile-fix from biondizzle/nvfp4-megamoe-kernel 2026-05-24 21:58:32 +00:00
biondizzle deleted tag v0.4-d1-hd256 from biondizzle/nvfp4-megamoe-kernel 2026-05-24 21:58:32 +00:00
biondizzle deleted tag d1.3-pre-sm100-helpers from biondizzle/nvfp4-megamoe-kernel 2026-05-24 21:58:31 +00:00
biondizzle deleted tag the-last-of-cutlass from biondizzle/nvfp4-megamoe-kernel 2026-05-24 21:58:31 +00:00
biondizzle deleted tag v-multitile-softmax-wip from biondizzle/nvfp4-megamoe-kernel 2026-05-24 21:58:31 +00:00
biondizzle deleted branch the-last-of-cutlass from biondizzle/nvfp4-megamoe-kernel 2026-05-24 21:57:17 +00:00
biondizzle deleted branch proper-nvfp4-integration from biondizzle/nvfp4-megamoe-kernel 2026-05-24 21:57:16 +00:00
biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-24 21:56:15 +00:00
77b366d44b Scrub B200 password from markdown files
83506e6ad2 Add MAY_24_26_PLAN.md: next session startup plan
9435bf9653 Restore NVFP4 Precision Roadmap + add O rescale gap to D1.5
03cbd8ffa6 Add STAGE_D2.md: Multi-query grid + head packing plan
f4e0315af9 Remove obsolete STAGE_D1.3.md and SMEM_P_GUIDANCE_REQUEST.md
Compare 641 commits »
biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-24 21:53:38 +00:00
27a69ca62b Scrub B200 password from markdown files
12091c2451 Add MAY_24_26_PLAN.md: next session startup plan
dcfc41b220 Restore NVFP4 Precision Roadmap + add O rescale gap to D1.5
bcfe52df0d Add STAGE_D2.md: Multi-query grid + head packing plan
05d45b9729 Remove obsolete STAGE_D1.3.md and SMEM_P_GUIDANCE_REQUEST.md
Compare 534 commits »
biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-24 21:50:34 +00:00
4108943590 Add MAY_24_26_PLAN.md: next session startup plan
biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-24 21:49:01 +00:00
f0d1bd0e18 Restore NVFP4 Precision Roadmap + add O rescale gap to D1.5