SolCoder

This model is a fine-tuned version of Pipper/SolCoder on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.5568

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: 0.0001
• train_batch_size: 37
• eval_batch_size: 37
• seed: 100
• distributed_type: multi-GPU
• num_devices: 4
• total_train_batch_size: 148
• total_eval_batch_size: 148
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss
0.6094	1.0	7440	0.6185
0.598	2.0	14880	0.6124
0.5845	3.0	22320	0.6075
0.5723	4.0	29760	0.6006
0.5589	5.0	37200	0.5943
0.5495	6.0	44640	0.5894
0.5371	7.0	52080	0.5861
0.5291	8.0	59520	0.5811
0.52	9.0	66960	0.5765
0.5095	10.0	74400	0.5746
0.5056	11.0	81840	0.5700
0.4967	12.0	89280	0.5682
0.4894	13.0	96720	0.5659
0.4861	14.0	104160	0.5619
0.4773	15.0	111600	0.5599
0.4754	16.0	119040	0.5599
0.4689	17.0	126480	0.5578
0.4642	18.0	133920	0.5575
0.4627	19.0	141360	0.5566
0.4573	20.0	148800	0.5568

Framework versions

• Transformers 4.33.0
• Pytorch 2.1.0+cu121
• Datasets 2.11.0
• Tokenizers 0.13.3