Instructions to use gopalanj/jevons-lfm25-1.2b-systemone with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- MLX
How to use gopalanj/jevons-lfm25-1.2b-systemone with MLX:
# Make sure mlx-lm is installed # pip install --upgrade mlx-lm # if on a CUDA device, also pip install mlx[cuda] # Generate text with mlx-lm from mlx_lm import load, generate model, tokenizer = load("gopalanj/jevons-lfm25-1.2b-systemone") prompt = "Once upon a time in" text = generate(model, tokenizer, prompt=prompt, verbose=True) - PEFT
How to use gopalanj/jevons-lfm25-1.2b-systemone with PEFT:
Task type is invalid.
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- LM Studio
- MLX LM
How to use gopalanj/jevons-lfm25-1.2b-systemone with MLX LM:
Generate or start a chat session
# Install MLX LM uv tool install mlx-lm # Generate some text mlx_lm.generate --model "gopalanj/jevons-lfm25-1.2b-systemone" --prompt "Once upon a time"
- Atomic Chat
Download hyperparams.json from gopalanj/jevons-lfm25-1.2b-systemone: direct link, hf CLI and curl.
- Browser
- Download file 2.49 kB
-
https://huggingface.co/gopalanj/jevons-lfm25-1.2b-systemone/resolve/main/hyperparams.json
- Command line
-
hf download hf://gopalanj/jevons-lfm25-1.2b-systemone/hyperparams.json
-
curl -L -o hyperparams.json https://huggingface.co/gopalanj/jevons-lfm25-1.2b-systemone/resolve/main/hyperparams.json
2.49 kB
| { | |
| "status": "promoted", | |
| "source_run": "adapters/lfm25-1.2b-systemone-seed", | |
| "model": "models/LFM2.5-1.2B-Instruct-8bit", | |
| "model_id": "lfm2.5-1.2b-instruct", | |
| "adapter_path": "adapters/lfm25-1.2b-systemone", | |
| "train_precision": "qlora-8bit", | |
| "rank": 16, | |
| "lora_scale": 2.0, | |
| "epochs": 2, | |
| "iters": 64, | |
| "batch_size": 4, | |
| "learning_rate": 5e-05, | |
| "num_layers": 16, | |
| "lora_keys": [ | |
| "self_attn.q_proj", | |
| "self_attn.k_proj", | |
| "self_attn.v_proj", | |
| "self_attn.out_proj", | |
| "feed_forward.w1", | |
| "feed_forward.w2", | |
| "feed_forward.w3" | |
| ], | |
| "mask_prompt": true, | |
| "choice_mode": "keys", | |
| "temperature": 1.0, | |
| "n_train_items": 78, | |
| "n_holdout_items": 17, | |
| "n_train_examples": 126, | |
| "n_holdout_examples": 29, | |
| "wall_seconds": 130.5, | |
| "val_nll": {"iter_1": 6.513, "iter_32": 3.628, "iter_64": 1.708}, | |
| "failed_grown_run": { | |
| "path": "adapters/lfm25-1.2b-systemone-grown", | |
| "n_train_items": 278, | |
| "n_train_examples": 476, | |
| "rank": 16, | |
| "lora_scale": 20.0, | |
| "epochs": 3, | |
| "wall_seconds": 534.1, | |
| "holdout_modal": 0.379, | |
| "note": "Repetitive grow.py templates + mlx-lm default scale 20 collapsed choice to sales/transfer/weather. Do not serve." | |
| }, | |
| "metrics_t1": { | |
| "holdout_base": { | |
| "n": 29, | |
| "schema_valid": 1.0, | |
| "modal_agreement": 0.724, | |
| "ece": 0.122, | |
| "brier": 0.390, | |
| "acc_at_conf_ge_0_8": 1.0, | |
| "n_conf_ge_0_8": 11, | |
| "by_type": {"choice": 0.769, "noul": 0.875, "score": 0.500} | |
| }, | |
| "holdout_lora": { | |
| "n": 29, | |
| "schema_valid": 1.0, | |
| "modal_agreement": 0.724, | |
| "ece": 0.248, | |
| "brier": 0.303, | |
| "acc_at_conf_ge_0_8": 1.0, | |
| "n_conf_ge_0_8": 14, | |
| "by_type": {"choice": 0.692, "noul": 0.750, "score": 0.750} | |
| }, | |
| "full_base": { | |
| "n": 155, | |
| "schema_valid": 1.0, | |
| "modal_agreement": 0.697, | |
| "ece": 0.107, | |
| "brier": 0.387, | |
| "acc_at_conf_ge_0_8": 0.921, | |
| "n_conf_ge_0_8": 63, | |
| "by_type": {"choice": 0.786, "noul": 0.682, "score": 0.561} | |
| }, | |
| "full_lora": { | |
| "n": 155, | |
| "schema_valid": 1.0, | |
| "modal_agreement": 0.839, | |
| "ece": 0.077, | |
| "brier": 0.193, | |
| "acc_at_conf_ge_0_8": 0.959, | |
| "n_conf_ge_0_8": 97, | |
| "by_type": {"choice": 0.857, "noul": 0.955, "score": 0.683} | |
| } | |
| }, | |
| "serve": "JEVONS_ADAPTER=adapters/lfm25-1.2b-systemone JEVONS_TEMPERATURE=1 JEVONS_CALIBRATION=off JEVONS_MODEL=models/LFM2.5-1.2B-Instruct-8bit uv run jevons serve" | |
| } | |