pythia-1_4b-deduped-measurement_pred-generated_stories

This model is a fine-tuned version of EleutherAI/pythia-1.4b-deduped on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.3621
• Accuracy: 0.8367
• Accuracy Sensor 0: 0.8459
• Auroc Sensor 0: 0.9242
• Accuracy Sensor 1: 0.8400
• Auroc Sensor 1: 0.9085
• Accuracy Sensor 2: 0.8252
• Auroc Sensor 2: 0.8916
• Accuracy Aggregated: 0.8356
• Auroc Aggregated: 0.9203

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: 1e-06
• train_batch_size: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 8
• num_epochs: 4
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Accuracy Sensor 0	Auroc Sensor 0	Accuracy Sensor 1	Auroc Sensor 1	Accuracy Sensor 2	Auroc Sensor 2	Accuracy Aggregated	Auroc Aggregated
No log	0.9948	119	0.5539	0.7015	0.7467	0.8498	0.6030	0.8246	0.7126	0.8175	0.7437	0.8571
0.6374	1.9979	239	0.3950	0.8215	0.8356	0.9138	0.8252	0.8929	0.8030	0.8800	0.8222	0.9106
0.4462	2.9927	358	0.3739	0.8285	0.8415	0.9228	0.8385	0.9077	0.8059	0.8863	0.8281	0.9176
0.3746	3.9791	476	0.3621	0.8367	0.8459	0.9242	0.8400	0.9085	0.8252	0.8916	0.8356	0.9203

Framework versions

• Transformers 4.41.0
• Pytorch 2.3.0+cu121
• Datasets 2.19.1
• Tokenizers 0.19.1