Upload SeRL checkpoints with ZipNN compression

README.md

@@ -0,0 +1,60 @@
+---
+license: apache-2.0
+tags:
+  - serl
+  - reinforcement-learning
+  - qwen2.5
+  - checkpoints
+---
+
+# SeRL Training Checkpoints
+
+Compressed checkpoints from SeRL (Self-Evolving Reinforcement Learning) experiments.
+Files use **ZipNN lossless compression** (~33% smaller, transparent loading).
+
+## Quick Start
+
+```bash
+pip install zipnn huggingface_hub transformers
+```
+
+```python
+# Enable ZipNN transparent loading
+from zipnn import zipnn_hf
+zipnn_hf()
+
+from huggingface_hub import snapshot_download
+from transformers import AutoModelForCausalLM, AutoTokenizer
+
+# Download specific checkpoint
+path = snapshot_download(
+    "AshwinKM2005/serl-checkpoints",
+    allow_patterns="serl_arce_qwen25_1_5b/huggingface/global_step200_hf/*"
+)
+
+# Load model (auto-decompresses .znn files)
+model = AutoModelForCausalLM.from_pretrained(
+    f"{path}/serl_arce_qwen25_1_5b/huggingface/global_step200_hf"
+)
+tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen2.5-1.5B")
+```
+
+## Experiments
+
+| Experiment | Base Model | Dataset |
+|------------|-----------|---------|
+| serl_arcc_qwen25_0_5b | Qwen2.5-0.5B | ARC-Challenge |
+| serl_ARC-c_qwen25_1_5b | Qwen2.5-1.5B | ARC-Challenge |
+| serl_arce_qwen25_0_5b | Qwen2.5-0.5B | ARC-Easy |
+| serl_arce_qwen25_1_5b | Qwen2.5-1.5B | ARC-Easy |
+
+## Structure
+
+Each experiment contains:
+- `huggingface/` - HuggingFace-compatible checkpoints
+- `deepspeed/` - DeepSpeed training checkpoints
+
+## Compression
+
+Files ending in `.safetensors.znn` are ZipNN compressed.
+The `zipnn_hf()` hook enables transparent loading without manual decompression.