diff options
| -rw-r--r-- | EMAIL_BEN_DRAFT2.md | 4 | ||||
| -rw-r--r-- | docs/campaign/CASCADE_ABLATION_PLAN.md | 7 |
2 files changed, 7 insertions, 4 deletions
diff --git a/EMAIL_BEN_DRAFT2.md b/EMAIL_BEN_DRAFT2.md index bb2674d..460c18b 100644 --- a/EMAIL_BEN_DRAFT2.md +++ b/EMAIL_BEN_DRAFT2.md @@ -45,8 +45,8 @@ Yuren 5. 成本表 2026-07-12 v2.1: 按当日查证实价（市场 H100 $1.87-2.99/h; AWS p5e blocks $4.97/GPU·h, 7月1日+20%后）+ 实测 amp 1.56×。旧 v1 锚 AWS-A100-savings 已废; 用户 H200 数据点按指示不进邮件, 仅作 $16.7/EF 最差上限留档。完整推导+价格出处: docs/campaign/COST_MODEL.md -6. **发送门: stage1b_amp (58.8k 全 epoch, GPU1) 完成且 ≈1.28 档才发**——邮件说了 "validated - against the fp32 reference this week"。若 amp 在 epoch 尺度掉链子 → 表格回退 TF32 (×1.5) 并删该句 +6. ~~发送门~~ **CLEARED (07-12 晚): stage1b_amp 终值 1.2868 vs fp32 1.2808 (Δ+0.006 噪声带内), + 零 guard,1.56× 全程成立**——已发邮件中 "validated against the fp32 reference this week" 完全闭合 --- # Draft 7 — reply to Ben's "LLM bullshit" Overleaf-comments email (2026-07-12 晚) diff --git a/docs/campaign/CASCADE_ABLATION_PLAN.md b/docs/campaign/CASCADE_ABLATION_PLAN.md index d604fc6..3377dab 100644 --- a/docs/campaign/CASCADE_ABLATION_PLAN.md +++ b/docs/campaign/CASCADE_ABLATION_PLAN.md @@ -578,8 +578,11 @@ quantization (wall-1: beta-displacement below bf16 resolution) — exactly as di cos 0.9999 at step 0. Zero guard events. - SPEED (solo GPU3/A6000, C512): amp 2.785 it/s vs fp32 1.789 it/s = 1.56x wall-clock; grows with width (tensor-core-bound share) -> treat 1.5x as the floor for 1-3B on H100. -- CAVEAT + confirm step: 4k is short-horizon; stage1b_amp (58.8k full epoch, identical recipe+seed, - GPU1) running as the epoch-scale confirm vs 1.2808. EMAIL_BEN_DRAFT2 send-gate #6 waits on it. +- ~~CAVEAT + confirm step~~ **EPOCH CONFIRM SEALED (2026-07-12 late): stage1b_amp DONE best val + CE 1.2868 vs fp32 1.2808 (Δ+0.006, inside the 0.02-0.03 best-of-noisy-val band); zero guard + events over 58.8k; 2.80 vs ~1.79 it/s = the 1.56x held for the full epoch.** amp = Stage-2 + default, full confidence. EMAIL_BEN_DRAFT2 gate #6 CLEARED (the sent "validated this week" + claim is now closed at epoch scale). - Cost consequence: COST_MODEL.md v2.1 (sourced July-2026 prices: market H100 $1.87-2.99/GPU.h, AWS p5e blocks $4.97/GPU.h post-hike) — with amp measured, 3B-Chinchilla ~$40k / 7Bx20B ~$30k on AWS blocks: BOTH inside the $50k envelope individually. amp is the Stage-2 default. |