Modelpublic

leobianco/bosch_RM_google_S130104_LLM_false_STRUCT_true_epo15_lr5.3e-04_r32_2609111447

sourceHugging Faceapache-2.0updated 16d agoView on Hugging Face

0likes137downloads

Model Card

boschRMgoogleS130104LLMfalseSTRUCTtrueepo15lr5.3e-04r32_2609111447

This model is a fine-tuned version of google/gemma-4-E4B-it on an unknown dataset. It achieves the following results on the evaluation set:

—Loss: 6.5124
—Roc Auc: 0.8173
—Best Threshold: 1.0000
—Tpr At Best Threshold: 0.9155
—Fpr At Best Threshold: 0.3182
—Accuracy At Best Threshold: 0.8602
—Avg Score True Positives: 1.0000
—Avg Score True Negatives: 0.9997

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

—learning_rate: 0.0005340062721199326
—trainbatchsize: 16
—evalbatchsize: 32
—seed: 130104
—distributed_type: multi-GPU
—num_devices: 2
—totaltrainbatch_size: 32
—totalevalbatch_size: 64
—optimizer: Use OptimizerNames.ADAMWTORCHFUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
—lrschedulertype: cosine
—lrschedulerwarmup_steps: 0.1
—num_epochs: 15.0

Training results

Training Loss	Epoch	Step	Validation Loss	Roc Auc	Best Threshold	Tpr At Best Threshold	Fpr At Best Threshold	Accuracy At Best Threshold	Avg Score True Positives	Avg Score True Negatives
No log	0	0	1.2476	0.4794	0.4526	0.9789	0.9091	0.7688	0.5883	0.5897
0.6362	1.1364	50	2.4158	0.6647	0.9929	0.8099	0.5	0.7366	0.9878	0.9502
0.2631	2.2727	100	2.2769	0.8035	0.9984	0.6408	0.1364	0.6935	0.9907	0.9530
0.4248	3.4091	150	2.1221	0.8818	0.9995	0.7676	0.0909	0.8011	0.9871	0.9175
0.1130	4.5455	200	4.0497	0.7997	1.0000	0.6549	0.1136	0.7097	0.9994	0.9979
0.8496	5.6818	250	3.9990	0.8600	0.9999	0.8169	0.2273	0.8065	0.9994	0.9785
0.0417	6.8182	300	6.2248	0.8119	1.0000	0.7676	0.1591	0.7849	0.9965	0.9895
0.1109	7.9545	350	4.2695	0.8925	1.0000	0.7394	0.0682	0.7849	0.9996	0.9847
0.0005	9.0909	400	5.6109	0.8505	1.0000	0.8451	0.2045	0.8333	1.0000	0.9992
0.0015	10.2273	450	5.6055	0.8600	1.0000	0.8521	0.2045	0.8387	1.0000	0.9989
0.0000	11.3636	500	6.1761	0.8271	1.0000	0.9014	0.2955	0.8548	1.0000	0.9995
0.0001	12.5	550	6.0289	0.8319	1.0	0.8592	0.25	0.8333	1.0000	0.9992
0.0030	13.6364	600	6.4768	0.8283	1.0	0.8803	0.2727	0.8441	1.0000	0.9996
0.0001	14.7727	650	6.5027	0.8182	1.0000	0.9155	0.3182	0.8602	1.0000	0.9997
0.0000	15.0	660	6.5124	0.8173	1.0000	0.9155	0.3182	0.8602	1.0000	0.9997

Framework versions

—PEFT 0.20.0
—Transformers 5.14.1
—Pytorch 2.11.0+cu130
—Datasets 5.0.1
—Tokenizers 0.22.2