Drop the training-step numbers from the model card

README.md

@@ -33,8 +33,7 @@ simulation environments live in the benchmark environment repository.
 
 Every method's weights are called **`ckpt`** whichever trainer wrote them — a
 directory for the JAX/orbax and LeRobot policies, a single file for Vid2Robot
-(`ckpt.pt`) and UniSkill (`ckpt.pth`). The training step each one corresponds
-to is in the table below, not in the path.
+(`ckpt.pt`) and UniSkill (`ckpt.pth`).
 
 `<condition>` is the training suite, and it is the **top** level: **`nd/`** = no
 distraction (clean scenes), **`wd/`** = with distraction (distractors in both
@@ -47,23 +46,13 @@ bank.
 
 ## Checkpoints
 
-| Paper label | Method dir | Weights | Step — `nd/` | Step — `wd/` | Size each |
-|---|---|---|---|---|---|
-| VLM + π0.5 | `vlm_pi05` | `ckpt/` | 50,000 | 50,000 | 9.35 GB |
-| Vid. Cond. π0.5 (Uniform) | `vid_cond_pi05_uniform` | `ckpt/` | 50,000 | 50,000 | 6.20 GB |
-| Vid. Cond. π0.5 (Keyframes) | `vid_cond_pi05_keyframes` | `ckpt/` | 50,000 | **20,000** | 6.20 GB |
-| Vid2Robot | `vid2robot` | `ckpt.pt` | 50,000 | 50,000 | 2.37 GB |
-| UniSkill | `uniskill_policy` | `ckpt.pth` | epoch 1,000 | epoch 1,000 | 0.74 GB |
-
-**The WD keyframes policy is from step 20,000, not 50,000.** That run reached
-20k steps, and it is the checkpoint the reported WD keyframe numbers come from.
-It is the one place where the two conditions are not trained equally long.
-
-The step directories these were renamed from, for cross-referencing against the
-training logs: `050000` (the LeRobot trainer saves at `num_train_steps`,
-zero-padded), `49999` and `20000` (openpi/orbax saves at
-`num_train_steps - 1`, unpadded) and `model_epoch_1000.pth` (Robomimic is
-epoch-based, 100 iterations per epoch).
+| Paper label | Method dir | Weights | Size each |
+|---|---|---|---|
+| VLM + π0.5 | `vlm_pi05` | `ckpt/` | 9.35 GB |
+| Vid. Cond. π0.5 (Uniform) | `vid_cond_pi05_uniform` | `ckpt/` | 6.20 GB |
+| Vid. Cond. π0.5 (Keyframes) | `vid_cond_pi05_keyframes` | `ckpt/` | 6.20 GB |
+| Vid2Robot | `vid2robot` | `ckpt.pt` | 2.37 GB |
+| UniSkill | `uniskill_policy` | `ckpt.pth` | 0.74 GB |
 
 UniSkill is a Robomimic diffusion policy over skill representations produced by
 an **off-the-shelf, un-finetuned UniSkill inverse-skill-dynamics model**. That