Text Classification
PEFT
Safetensors
English
Japanese
d1a
decision-model
calibration
lora
gemma4
typesafe
system-one
pull-requests
on-device
Instructions to use JohnP1/d1a-e2b with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- PEFT
How to use JohnP1/d1a-e2b with PEFT:
from peft import PeftModel from transformers import AutoModel base_model = AutoModel.from_pretrained("google/gemma-4-E2B") model = PeftModel.from_pretrained(base_model, "JohnP1/d1a-e2b") - Notebooks
- Google Colab
- Kaggle
Download training_config.json from JohnP1/d1a-e2b: direct link, hf CLI and curl.
- Browser
- Download file 2.09 kB
-
https://huggingface.co/JohnP1/d1a-e2b/resolve/main/training_config.json
- Command line
-
hf download hf://JohnP1/d1a-e2b/training_config.json
-
curl -L -o training_config.json https://huggingface.co/JohnP1/d1a-e2b/resolve/main/training_config.json
2.09 kB
| { | |
| "args": { | |
| "base": "google/gemma-4-E2B", | |
| "n_per_source": 1000, | |
| "epochs": 1, | |
| "lr": 5e-05, | |
| "head_lr": 0.0, | |
| "weight_decay": 0.01, | |
| "lora": 16, | |
| "accum": 4, | |
| "holdout": "", | |
| "suite": "evals/v7/decision-v7", | |
| "train_sources": "", | |
| "device": "cuda", | |
| "batch": 2, | |
| "dtype": "bf16", | |
| "weights_dtype": "bf16", | |
| "checkpointing": 1, | |
| "head_dim": 256, | |
| "lora_targets": "all", | |
| "base_revision": "d29ff6b45f081a49ee2733a859c9c9c2d95d1a6f", | |
| "p_none": 0.1, | |
| "p_none_distract": 0.12, | |
| "p_distract": 0.15, | |
| "p_none_pair": 0.0, | |
| "none_pair_max_state": null, | |
| "synthetic_repeat": 1, | |
| "public_frac": 1.0, | |
| "out": "e2b-v06", | |
| "data": "data/train-e2b-v06.jsonl", | |
| "max_state": 5120, | |
| "extra_suites": "", | |
| "replay": 0, | |
| "init_from": "JohnP1/d1a-e2b@v0.2.1-2epoch-calibrated", | |
| "allow_missing_sources": "", | |
| "reason": "", | |
| "row_budget": 0, | |
| "shared_prefix": 0, | |
| "length_sort": 0, | |
| "pass_tokens_max": 0, | |
| "max_steps": 0, | |
| "save_every_steps": 0, | |
| "save_every_minutes": 30.0, | |
| "resume": 1, | |
| "stop_after": 0, | |
| "seed": 0 | |
| }, | |
| "suite_sha256": "a8f50e481b7d90b97da049e0ff6a01cee2f1ed204aed61a8265af0edbb5514d2", | |
| "base_revision": "d29ff6b45f081a49ee2733a859c9c9c2d95d1a6f", | |
| "init_source": { | |
| "init_from": "JohnP1/d1a-e2b@v0.2.1-2epoch-calibrated", | |
| "resolved": "JohnP1/d1a-e2b@b529d2fd25bb813319903ff8f1304e670677a6bb", | |
| "weights_sha256": "e23dbe51453b6e926da0db70a38d680bc64644caff872cf0defcbbd8b1acbe46", | |
| "head_sha256": "fb3173876e55edb96bc73fab852372f2e4f6a193e346d6cd80868c8bbdda6990", | |
| "tensors": 410 | |
| }, | |
| "holdout": [], | |
| "sources": { | |
| "covered": [ | |
| "evals/hard-v1", | |
| "evals/devtools-v1", | |
| "evals/documents-v1", | |
| "evals/d1a/ja-jglue:train", | |
| "evals/d1a/pr-labels:train", | |
| "evals/d1a/pr-labels:train-ja", | |
| "evals/d1a/pr-labels:train-blast", | |
| "evals/d1a/routing:factory-train", | |
| "evals/d1a/routing:generic-train", | |
| "evals/v7/decision-v7" | |
| ], | |
| "allowed_missing": [], | |
| "reason": null | |
| } | |
| } | |