Views
No views yet
| Parameter | Value |
|---|---|
| data_path | /home/felipe/hook_prompt_enhancer/training/data/skeleton_train.jsonl |
| data_splits | {'train': 540, 'val': 60} |
| hyperparameters | {'learning_rate': 0.001, 'batch_size': 4, 'epochs': 3, 'lora_r': 64, 'lora_alpha': 128, 'lora_dropout': 0.05, 'gradient_accumulation_steps': 1, 'weight_decay': 0.01, 'warmup_ratio': None, 'time_budget': 300} |
| model_name | Qwen/Qwen2.5-1.5B-Instruct |
| seed | 42 |
| stage | skeleton |
| timestamp | 2026-03-28T10:13:10.516948+00:00 |
| val_loss | 0.4951 |
1from llama_cpp import Llama
2
3llm = Llama(
4 model_path="qwen2.5-1.5b-instruct-q4_k_m.gguf",
5 lora_path="skeleton-lora.gguf",
6)
7output = llm("Decompose this prompt: ...")