model_20m_11B card

model_20m_11B/README.md

@@ -0,0 +1,42 @@
+---
+license: apache-2.0
+tags: [chess, reinforcement-learning, grpo]
+---
+
+# model_20m_11B — RL (GRPO) checkpoints
+
+RL post-training trajectory for the chess pre-to-post compute-allocation study.
+The pretraining base and SFT init for this model are
+[`model_20m_11B`](https://huggingface.co/pavelslab-nyu/Chess-Pretrain-Models/tree/main/model_20m_11B) and
+[`model_20m_11B`](https://huggingface.co/pavelslab-nyu/Chess-SFT-Models/tree/main/model_20m_11B).
+
+| | |
+|---|---|
+| run id | `C6p5e18_20m_alpha0.200_beta0.008` |
+| parameters | 20m |
+| pretraining tokens | 10,549,361,403 (10.5B) |
+| compute class | 6p5e18 |
+| alpha (pretrain fraction) | 0.2 |
+| beta (SFT fraction) | 0.008 |
+| checkpoints here | 50 (steps 100–5000) |
+| checkpoints saved by the run | 100 |
+
+## Steps
+
+`100`, `200`, `300`, `400`, `500`, `600`, `700`, `800`, `900`, `1000`, `1100`, `1200`, `1300`, `1400`, `1500`, `1600`, `1700`, `1800`, `1900`, `2000`, `2100`, `2200`, `2300`, `2400`, `2500`, `2600`, `2700`, `2800`, `2900`, `3000`, `3100`, `3200`, `3300`, `3400`, `3500`, `3600`, `3700`, `3800`, `3900`, `4000`, `4100`, `4200`, `4300`, `4400`, `4500`, `4600`, `4700`, `4800`, `4900`, `5000`
+
+## Loading
+
+Each `global_step_N/` folder is self-contained. The models use a custom
+tokenizer (`tokenizer.py`), and the remote-code resolver ignores `subfolder=`,
+so download the folder first and load the **local path**:
+
+```python
+from huggingface_hub import snapshot_download
+from transformers import AutoModelForCausalLM, AutoTokenizer
+
+step = "global_step_5000"
+p = snapshot_download("Pre2Post-Chess-RL/Chess-RL-Models", allow_patterns=f"model_20m_11B/{step}/*") + f"/model_20m_11B/{step}"
+model = AutoModelForCausalLM.from_pretrained(p, trust_remote_code=True)
+tok = AutoTokenizer.from_pretrained(p, trust_remote_code=True)
+```