model card

README.md

@@ -22,20 +22,21 @@ point in any eval CSV maps to a folder here without a lookup table:
 
 ```
 <model>_mbpp_<arm>_<lr>/
-    global_step_<N>/     a complete HF model directory (stride 30 + endpoint)
+    global_step_<N>/     a complete HF model directory
     logs/                trainer logs
     rollouts/            sampled rollouts
 ```
 
 | folder | base model | arm | lr | checkpoints |
 | --- | --- | --- | --- | --- |
-| `qwen3b_mbpp_hardcd_lam0p1_lr6e7` | `Qwen/Qwen2.5-3B` | replay buffer (lambda=0.1) | 6e-7 | 30-800 step 30 |
-| `qwen3b_mbpp_nobuf_lr6e7` | `Qwen/Qwen2.5-3B` | no buffer | 6e-7 | 330-800 step 30 |
+| `qwen3_1p7b_base_mbpp_nobuf_lr6e7` | `Qwen/Qwen3-1.7B-Base` | no buffer | 6e-7 | 30-980 46 ckpts, irregular |
+| `qwen3b_mbpp_hardcd_lam0p1_lr6e7` | `Qwen/Qwen2.5-3B` | replay buffer (lambda=0.1) | 6e-7 | 30-800 27 ckpts, irregular |
+| `qwen3b_mbpp_nobuf_lr6e7` | `Qwen/Qwen2.5-3B` | no buffer | 6e-7 | 30-800 27 ckpts, irregular |
 
 ```python
 from transformers import AutoModelForCausalLM
 m = AutoModelForCausalLM.from_pretrained(
-        "RL-Forgettting-Exp4/code-rl-lr-ablation", subfolder="qwen3b_mbpp_hardcd_lam0p1_lr6e7/global_step_800")
+        "RL-Forgettting-Exp4/code-rl-lr-ablation", subfolder="qwen3_1p7b_base_mbpp_nobuf_lr6e7/global_step_980")
 ```
 
 ## Training