distilgpt2-mentalchat16k

This model is a fine-tuned version of distilgpt2 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 1.7166

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: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 8
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.03
• num_epochs: 4
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
2.6182	0.1496	100	2.2105
2.2076	0.2992	200	2.0442
2.0966	0.4488	300	1.9762
2.0588	0.5984	400	1.9349
2.0266	0.7479	500	1.8990
1.9837	0.8975	600	1.8681
1.955	1.0471	700	1.8443
1.9298	1.1967	800	1.8224
1.9156	1.3463	900	1.8075
1.9128	1.4959	1000	1.7937
1.9054	1.6455	1100	1.7831
1.8468	1.7951	1200	1.7702
1.8663	1.9447	1300	1.7631
1.8413	2.0942	1400	1.7529
1.8356	2.2438	1500	1.7486
1.8324	2.3934	1600	1.7444
1.8652	2.5430	1700	1.7358
1.8409	2.6926	1800	1.7331
1.8162	2.8422	1900	1.7313
1.8211	2.9918	2000	1.7283
1.828	3.1414	2100	1.7262
1.8139	3.2909	2200	1.7247
1.8168	3.4405	2300	1.7251
1.8191	3.5901	2400	1.7246
1.8285	3.7397	2500	1.7235
1.8085	3.8893	2600	1.7237

Framework versions

• PEFT 0.13.2
• Transformers 4.46.3
• Pytorch 2.4.1+cu118
• Datasets 3.1.0
• Tokenizers 0.20.3