biondizzle

0 Followers · 0 Following

• Joined on 2025-12-10

Repositories

25

Projects Packages Public Activity Starred Repositories

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-23 00:30:50 +00:00

a63f452c86 Fix softmax loop: use self.n_kv_tiles not cute.size(gK, mode=[3])

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-23 00:29:54 +00:00

195b0506af auto: pre-test commit

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-23 00:28:46 +00:00

88b66e9dca Add O rescale with pre-built paired atoms (corr_tile_size=16)

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-23 00:27:43 +00:00

fae9f6fbb5 Reset to working_softmax_maybe.py + TMA fix only

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-23 00:25:20 +00:00

524b5b1840 Fix final normalize: use working 2D register tensor pattern from working_softmax_maybe.py

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-23 00:23:40 +00:00

85973743d6 Fix: add self.n_kv_tiles to __init__

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-23 00:22:15 +00:00

261f23e698 Add per-tile O rescale (O *= acc_scale) to softmax loop

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-23 00:17:10 +00:00

524f0bdfb4 Clean up: archive diagnostics and superseded tests

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-23 00:05:08 +00:00

8273a94506 auto: pre-test commit

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-23 00:02:35 +00:00

3a6f946316 auto: pre-test commit

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-23 00:00:48 +00:00

9b51586742 auto: pre-test commit

biondizzle pushed tag v-tma-multitile-fix to biondizzle/nvfp4-megamoe-kernel 2026-05-22 23:51:34 +00:00

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-22 23:51:31 +00:00

89b86c4134 🚀🚀🚀 TMA MULTI-TILE FIX VERIFIED ON B200 🚀🚀🚀

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-22 23:35:56 +00:00

6db7fd339d FIX: (None,0,None,0) for ALL tma_partition outputs — verified shapes on B200

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-22 23:34:05 +00:00

a50cb138c8 auto: pre-test commit

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-22 23:30:45 +00:00

e3054b5937 auto: pre-test commit

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-22 23:29:16 +00:00

9a3cf248db auto: pre-test commit

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-22 23:28:19 +00:00

8718e82258 auto: pre-test commit

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-22 23:27:34 +00:00

d45b6173f5 auto: pre-test commit

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-22 23:25:41 +00:00

4ce5926498 FIX: (None,0,None,0) pre-slice keeps KV tile axis (mode 2) free