Upload QUICK_REFERENCE.md with huggingface_hub

QUICK_REFERENCE.md

@@ -0,0 +1,139 @@
+# 🚀 CodeLlama Migration - Quick Reference Guide
+
+**Last Updated:** 2025-11-25 05:55 UTC
+
+---
+
+## 📁 Key Paths
+
+### **Base Model**
+```
+codellama-migration/models/base-models/CodeLlama-7B-Instruct/
+```
+**Status:** ⏳ Downloading (~10-15 minutes)
+
+### **Processed Dataset**
+```
+codellama-migration/datasets/processed/elinnos_fifo_codellama_v1.jsonl
+```
+**Status:** ✅ Ready (94 samples)
+
+### **Training Output Directory**
+```
+codellama-migration/training-outputs/
+```
+**Status:** ⏳ Waiting for training
+
+### **Updated Inference Script**
+```
+codellama-migration/scripts/inference/inference_codellama.py
+```
+**Status:** ✅ Updated with code extraction
+
+### **Progress Tracker**
+```
+codellama-migration/MIGRATION_PROGRESS.md
+```
+**Status:** ✅ Updated in real-time
+
+---
+
+## 🔧 Training Command (When Ready)
+
+```bash
+cd /workspace/ftt
+
+# Split dataset first
+python3 -c "
+import json
+import random
+random.seed(42)
+samples = [json.loads(l) for l in open('codellama-migration/datasets/processed/elinnos_fifo_codellama_v1.jsonl')]
+random.shuffle(samples)
+train = samples[:75]
+with open('codellama-migration/datasets/processed/train.jsonl', 'w') as f:
+    for s in train:
+        f.write(json.dumps(s) + '\n')
+"
+
+# Training command (adjust paths when model is downloaded)
+cd semicon-finetuning-scripts
+python3 models/msp/ft/finetune_mistral7b.py \
+    --base-model /workspace/ftt/codellama-migration/models/base-models/CodeLlama-7B-Instruct \
+    --dataset /workspace/ftt/codellama-migration/datasets/processed/train.jsonl \
+    --output-dir /workspace/ftt/codellama-migration/training-outputs/mistral-finetuned-codellama-v1 \
+    --max-length 2048
+```
+
+**Recommended Parameters:**
+- Epochs: 5 (instead of 3)
+- Learning Rate: 2e-5 (instead of 5e-5)
+- LoRA Rank: 64 (instead of 32)
+- LoRA Alpha: 128 (instead of 64)
+
+---
+
+## 📊 Training Parameters Reference
+
+| Parameter | Old Value | New Value |
+|-----------|-----------|-----------|
+| **Epochs** | 3 | **5** |
+| **Learning Rate** | 5e-5 | **2e-5** |
+| **LoRA Rank** | 32 | **64** |
+| **LoRA Alpha** | 64 | **128** |
+| **Temperature** | 0.7 | **0.3** |
+
+---
+
+## 🔍 Monitoring Downloads
+
+```bash
+# Check download progress
+tail -f codellama-migration/download_log.txt
+
+# Check if download is complete
+ls -lh codellama-migration/models/base-models/CodeLlama-7B-Instruct/
+
+# Expected files when complete:
+# - config.json
+# - tokenizer.json
+# - tokenizer_config.json
+# - pytorch_model-*.bin (or .safetensors)
+```
+
+---
+
+## ✅ Completed Tasks Checklist
+
+- [x] Folder structure created
+- [x] Dataset reformatted (94 samples)
+- [x] Inference script updated
+- [x] Training script symlinks created
+- [x] Progress tracker created
+- [ ] Model downloaded (in progress)
+- [ ] Dataset split (train/val/test)
+- [ ] Training completed
+- [ ] Testing completed
+
+---
+
+## 🎯 Next Steps
+
+1. **Wait for model download** (~10-15 minutes)
+   - Monitor: `tail -f codellama-migration/download_log.txt`
+   
+2. **Split dataset** into train/val/test
+   - 75 training / 9 validation / 10 test
+   
+3. **Start training** with CodeLlama
+   - Use updated parameters
+   - Output to `codellama-migration/training-outputs/`
+   
+4. **Test** on 3 training + 3 test samples
+   - Compare with previous Mistral results
+
+---
+
+**For detailed progress, see:** `MIGRATION_PROGRESS.md`
+
+