biondizzle

biondizzle pushed to mega-moe-nvfp4 at biondizzle/deepseek-v4-quant

2026-05-11 20:29:38 +00:00

7a4403fa98 feat: FP4 staging kernel - BF16 → E2M1 packed + UE4M3 block16 scales

biondizzle pushed to nvfp4-mega-moe at biondizzle/DeepGEMM

2026-05-11 20:29:09 +00:00

b3d1aae038 feat: full FP4 activations for mxf4nvf4 - E2M1 packed A side + UE4M3 scales

biondizzle pushed to mega-moe-nvfp4 at biondizzle/deepseek-v4-quant

2026-05-11 19:55:43 +00:00

0fd2d4f078 diag: add weight_scale uint8 histogram to verify E8M0 vs E4M3 format

biondizzle pushed to mega-moe-nvfp4 at biondizzle/deepseek-v4-quant

2026-05-11 19:51:50 +00:00

50a945bde4 bump cache buster 29

biondizzle pushed to mega-moe-nvfp4 at biondizzle/deepseek-v4-quant

2026-05-11 19:51:47 +00:00

48b905406a diag: add CUDA sync after mega_moe finalize + forward to catch errors

biondizzle pushed to mega-moe-nvfp4 at biondizzle/deepseek-v4-quant

2026-05-11 19:40:10 +00:00

35f6b66678 fix: UE8M0 reinterpret in DeepGEMM fold_global_scale + bump cache

biondizzle pushed to nvfp4-mega-moe at biondizzle/DeepGEMM

2026-05-11 19:40:06 +00:00

2cd86ff5e7 fix: UE8M0→float32 reinterpret in fold_global_scale (Bug #7)

biondizzle pushed to mega-moe-nvfp4 at biondizzle/deepseek-v4-quant

2026-05-11 19:26:23 +00:00

f32d6b5b48 bump cache buster to 27

biondizzle pushed to mega-moe-nvfp4 at biondizzle/deepseek-v4-quant

2026-05-11 19:26:14 +00:00

cd24182e36 diag: add NaN/Inf + FP8-dtype checks after NVFP4 dequant

biondizzle pushed to mega-moe-nvfp4 at biondizzle/deepseek-v4-quant

2026-05-11 18:48:08 +00:00

8ae2214bad fix: reorder Dockerfile ARG before COPY for proper cache busting

biondizzle pushed to mega-moe-nvfp4 at biondizzle/deepseek-v4-quant

2026-05-11 16:38:54 +00:00

c4891e9ee2 fix: manual FP32→UE4M3 quant in Triton staging kernel

biondizzle pushed to nvfp4-mega-moe at biondizzle/DeepGEMM

2026-05-11 16:25:10 +00:00

47621bb990 add NVFP4SymmBuffer + get_symm_buffer_for_nvfp4_mega_moe Python wrapper

biondizzle pushed to mega-moe-nvfp4 at biondizzle/deepseek-v4-quant

2026-05-11 16:12:57 +00:00

436109081c bump cache buster to 24

biondizzle pushed to mega-moe-nvfp4 at biondizzle/deepseek-v4-quant

2026-05-11 16:12:40 +00:00

5faf9916eb fix: UE4M3 activation scales + group_size=16 for NVFP4 mega_moe

biondizzle pushed to nvfp4-mega-moe at biondizzle/DeepGEMM

2026-05-11 16:11:14 +00:00

86a1263f44 fix: gran_k=16 in transform_sf + sm_100a arch for NVFP4 mega_moe

biondizzle pushed to nvfp4-mega-moe at biondizzle/DeepGEMM

2026-05-11 15:04:50 +00:00

fbdddaccf4 revert: restore mxf4nvf4/block16 code (correct path for sm_100a)

biondizzle pushed to nvfp4-mega-moe at biondizzle/DeepGEMM

2026-05-11 14:24:56 +00:00

e80fe9af60 docs: CORRECTED — mxf4nvf4 IS supported on sm_100a (B200)

biondizzle pushed to mega-moe-nvfp4 at biondizzle/deepseek-v4-quant

2026-05-11 14:24:14 +00:00

220649c188 docs: CORRECTED — mxf4nvf4 IS supported on sm_100a (B200)

biondizzle pushed to nvfp4-mega-moe at biondizzle/DeepGEMM

2026-05-11 13:55:19 +00:00

c2f4a30780 docs: comprehensive README update through build 22

biondizzle pushed to mega-moe-nvfp4 at biondizzle/deepseek-v4-quant

2026-05-11 13:53:42 +00:00

cfead0012d docs: comprehensive README update through build 22