lemexp-hol-thms-by-file-deepseek-coder-1.3b-base

This model is a fine-tuned version of deepseek-ai/deepseek-coder-1.3b-base on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.2839

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: 0.0002
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 100
• num_epochs: 6
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
0.4174	0.2000	3777	0.4078
0.34	0.4000	7554	0.3589
0.3104	0.6001	11331	0.3305
0.2901	0.8001	15108	0.3113
0.2823	1.0001	18885	0.2998
0.2591	1.2001	22662	0.2958
0.2597	1.4001	26439	0.2871
0.2536	1.6002	30216	0.2895
0.2353	1.8002	33993	0.2819
0.2236	2.0002	37770	0.2784
0.2055	2.2002	41547	0.2898
0.2129	2.4003	45324	0.2687
0.2001	2.6003	49101	0.2719
0.2108	2.8003	52878	0.2714
0.2023	3.0003	56655	0.2650
0.1811	3.2003	60432	0.2709
0.1702	3.4004	64209	0.2655
0.176	3.6004	67986	0.2665
0.1702	3.8004	71763	0.2666
0.1597	4.0004	75540	0.2620
0.1388	4.2004	79317	0.2704
0.1452	4.4005	83094	0.2707
0.146	4.6005	86871	0.2705
0.1388	4.8005	90648	0.2622
0.1435	5.0005	94425	0.2649
0.1295	5.2006	98202	0.2794
0.1195	5.4006	101979	0.2780
0.124	5.6006	105756	0.2796
0.1183	5.8006	109533	0.2839

Framework versions

• PEFT 0.14.0
• Transformers 4.47.0
• Pytorch 2.5.1+cu124
• Datasets 3.2.0
• Tokenizers 0.21.0