Update README.md

README.md

@@ -13,7 +13,7 @@ tags:
   - wan2.1
 ---
 
-# LiveWan — streaming text-to-video, step 2250
+# LiveWan — streaming text 2 video, 3000 steps
 
 > **Unofficial community project.** Not affiliated with, endorsed by, or produced
 > by Alibaba Group or the Wan-Video team. Built on their Apache-2.0
@@ -49,7 +49,7 @@ sentence.
 | `out/world_p{0,44,60,82}.pt` | 4 × 59 MB | the four cached evaluation worlds — skip base-model generation entirely |
 | `samples/` | 16 MB | reference clips and analysis filmstrips from those worlds |
 
-**To continue the run — a further 28.3 GB.** Not needed for inference; do not
+**To continue the run** Not needed for inference. do not
 download these unless you intend to train.
 
 | path | size | what |
@@ -90,44 +90,14 @@ browser stream.
 **`--prompt-idx` must match the world**: `world_p60.pt` goes with `--prompt-idx 60`.
 Indices run 0–95; anything above fails.
 
-## Why step 2250 and not the final step 3000
-
-The run went to 3000 iterations. Step 3000 is **not** the released checkpoint: it
-more than doubles the sharpness-decay deviation against 2250, almost entirely on
-world 60, where sharpness *rises* 30% across a 30-second stream.
-
-| arm | step | mean \|decay−1\| | mean \|drift\| | mean motion | min real-time |
-|---|---|---|---|---|---|
-| base | 2000 | 0.0775 | 0.0160 | 1.05 | 2.72× |
-| prior run | 400 | 0.1325 | 0.0555 | 2.69 | 2.75× |
-| this run | 750 | 0.0750 | 0.0915 | 1.91 | 2.49× |
-| this run | 1500 | 0.0725 | 0.0580 | 1.93 | 2.43× |
-| **this run** | **2250** | **0.0575** | **0.0178** | 2.17 | 2.77× |
-| this run | 3000 | 0.1275 | 0.0240 | 2.21 | 2.76× |
-
-(Real-time column as originally published, against a 25 fps budget; against the
-native 16 fps it is 3.8–4.4×.)
-
-Checkpoint quality here is **not monotone** — an earlier run went −0.024 → −0.256 →
-−0.574 → −0.115 on drift across four consecutive checkpoints. The honest
-counter-argument for step 3000, which is sharper and has more motion, is in the
-code repo's `docs/HANDOFF.md` §1.
-
-Step 3000 ships here as `checkpoints/t14b_b64/latest.pt` so the comparison is
-yours to make rather than to take on trust. Swap it into the command above and
-watch world 60 — that is where the two checkpoints actually differ.
-
-`_noema` means the EMA copy was stripped to halve the file, 10.6 → 5.3 GB. This
-changes nothing for inference: every evaluation in this project ran with
-`use_ema: false`.
 
 ## Training
 
 3000 iterations of SF-DMD distillation from a Wan2.1-T2V-14B teacher into a
 Wan2.1-T2V-1.3B student, effective batch 64 (8×H200, accum 8, FSDP), **41.6 hours**
 at 66.5 s/it, zero interventions. Losses do not decrease in this trainer and should
-not — the critic is retrained every step, so the generator holds position against a
-strengthening opponent.
+not (the critic is retrained every step, so the generator holds position against a
+strengthening opponent).
 
 ## Continuing the run