Instructions to use enzinol/Fred with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- PEFT
How to use enzinol/Fred with PEFT:
from peft import PeftModel from transformers import AutoModel base_model = AutoModel.from_pretrained("LiquidAI/LFM2.5-350M") model = PeftModel.from_pretrained(base_model, "enzinol/Fred") - Notebooks
- Google Colab
- Kaggle
Download training_config.json from enzinol/Fred: direct link, hf CLI and curl.
- Browser
- Download file 1.21 kB
-
https://huggingface.co/enzinol/Fred/resolve/main/training_config.json
- Command line
-
hf download hf://enzinol/Fred/training_config.json
-
curl -L -o training_config.json https://huggingface.co/enzinol/Fred/resolve/main/training_config.json
1.21 kB
| { | |
| "args": { | |
| "base": "LiquidAI/LFM2.5-350M", | |
| "n_per_source": 1000, | |
| "epochs": 2, | |
| "lr": 0.0002, | |
| "head_lr": 0.0, | |
| "weight_decay": 0.01, | |
| "lora": 16, | |
| "accum": 8, | |
| "holdout": "", | |
| "perm_kl": 0.1, | |
| "perm_frac": 0.3, | |
| "ord_w": 0.0, | |
| "suite": "train/lev-suite", | |
| "train_sources": "", | |
| "device": "mps", | |
| "batch": 1, | |
| "dtype": "fp32", | |
| "checkpointing": 0, | |
| "option_isolation": 0, | |
| "special_embeddings": 0, | |
| "head_dim": 256, | |
| "lora_targets": "all", | |
| "base_revision": "", | |
| "p_none": 0.1, | |
| "p_none_distract": 0.12, | |
| "p_distract": 0.15, | |
| "p_none_pair": 0.0, | |
| "synthetic_repeat": 1, | |
| "public_frac": 1.0, | |
| "anchor": "", | |
| "anchor_w": 0.0, | |
| "anchor_sources": "", | |
| "p_subset": 0.0, | |
| "subset_none": 0.7, | |
| "subset_absent": 0.15, | |
| "bucket_window": 0, | |
| "empty_cache_every": 1, | |
| "save_every": 0, | |
| "resume": false, | |
| "stop_after_step": 0, | |
| "out": "runs/lev350-hard", | |
| "seed": 0 | |
| }, | |
| "suite_sha256": "8463884553efe0ee2fd9037c9f8893f2a6cee62c2beec90d5c967dea375bec06", | |
| "base_revision": "9e6c6ccf47cd318696e137d381a7ded8fe4df09f", | |
| "ordinal_objective": "ranked_probability_score", | |
| "holdout": [] | |
| } | |