biondizzle

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel

2026-06-02 21:20:45 +00:00

0b6ca0df80 P5 integration + B3 q_a_norm fused + gsa scalar fix

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel

2026-06-02 20:23:49 +00:00

7e42b5e090 A1: Add ◇ (think_start) priming after Assistant token

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel

2026-06-02 20:16:51 +00:00

ac4eedc444 auto: pre-test commit

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel

2026-06-02 19:59:54 +00:00

ecd48ab65e A1: Add explicit stop set for DSV4 turn-end tokens

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel

2026-06-02 19:27:49 +00:00

35dbb8d12b Cleanup Part 2: Fix docs, stale references, dead code

f3b551956d Cleanup Step 2: Archive Lineage P code, fix broken imports

8de47e26ce Cleanup Step 1: Move root-level files to proper directories

Compare 3 commits »

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel

2026-06-02 19:08:53 +00:00

b111525af4 Fix indexer documentation and safety issues

d770111cb1 Remove stale duplicate .cu files from indexer/ subfolder

Compare 2 commits »

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel

2026-06-02 18:49:34 +00:00

eb5ef93bf1 Add A/B comparison mode for P4 fused vs unfused RMSNorm+quantize

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel

2026-06-02 18:26:04 +00:00

b8bab01a55 Update PERFORMANCE_AUDIT.md — P4 done, P5 kernel done (pending integration)

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel

2026-06-02 18:12:00 +00:00

8447ba7138 FIX: Deadlock in indexer_score_topk kernel — __syncthreads inside strided loop

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel

2026-06-02 17:57:34 +00:00

c926c4a597 P5: Fix mhc_rmsnorm_quantize_nvfp4 — add proper function definition

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel

2026-06-02 17:52:06 +00:00

36fdbeb56d stuff

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel

2026-06-02 17:43:28 +00:00

bdf0b15d45 P4: Fix rmsnorm_quantize_nvfp4 returns QuantizedActivation not tuple

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel

2026-06-02 16:39:44 +00:00

454dbdad52 P5: Fused mHC pre_block + RMSNorm + NVFP4 quantize kernel

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel

2026-06-02 16:38:45 +00:00

7bb3207347 P4: Integrate fused RMSNorm+quantize into single_shot (attention path)

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel

2026-06-02 16:37:39 +00:00

0d1cd1e216 P4: Add QuantizedActivation + Nvfp4Linear.run_from_quantized

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel

2026-06-02 16:34:52 +00:00

149ecefb56 P4: Relax test thresholds — per-row gsa vs scalar gsa difference expected

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel

2026-06-02 16:31:57 +00:00

57ab4b9d4c P4: Fix dequantize_nvfp4 bridge — handle float8_e4m3fn dtype

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel

2026-06-02 16:28:46 +00:00

29f836d711 P4: Fix fused RMSNorm kernel — match quantize_nvfp4.cu encoding

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel

2026-06-02 16:26:29 +00:00

794ebaf7e5 P4: Fused RMSNorm + NVFP4 quantize kernel (2 launches vs 6+)

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel

2026-06-02 11:19:07 +00:00

82294fc21e Fix nope_dim UnboundLocalError — hoist to function scope