| library_name: peft | |
| base_model: meta-llama/Llama-2-7b-hf | |
| tags: | |
| - lora | |
| - llama2 | |
| - mmlu | |
| - data-selection | |
| # full-sft | |
| Offline training on the **full dataset** (100%) with BM25-retrieved data. | |
| > **Note**: This checkpoint is from a **single random seed** (seed=3) and a specific training step (step 4220). Results may vary across seeds. | |
| ## Details | |
| | Key | Value | | |
| |-----|-------| | |
| | Base model | `meta-llama/Llama-2-7b-hf` | | |
| | Task | MMLU | | |
| | Data selection | Full SFT (BM25) | | |
| | Data ratio | 100% | | |
| | Online | False | | |
| | LoRA rank | 128 | | |
| | LoRA alpha | 512 | | |
| | Target modules | q_proj, k_proj, v_proj, o_proj | | |
| | Seed | 3 | | |
| | Checkpoint step | 4220 | | |
| ## Usage | |
| ```python | |
| from peft import PeftModel | |
| from transformers import AutoModelForCausalLM, AutoTokenizer | |
| base_model = AutoModelForCausalLM.from_pretrained("meta-llama/Llama-2-7b-hf") | |
| model = PeftModel.from_pretrained(base_model, "DATA-ADAPT/full-sft") | |
| tokenizer = AutoTokenizer.from_pretrained("DATA-ADAPT/full-sft") | |
| ``` | |