falcon-7b-sharded-bf16-finetuned-mental-health-dsm5mistral

This model is a fine-tuned version of ybelkada/falcon-7b-sharded-bf16 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 1.7448

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
• gradient_accumulation_steps: 4
• total_train_batch_size: 64
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.03
• training_steps: 200

Training results

Training Loss	Epoch	Step	Validation Loss
2.1353	0.1028	10	2.2214
2.2465	0.2057	20	2.1292
1.8446	0.3085	30	2.0584
1.9796	0.4113	40	1.9319
1.6682	0.5141	50	2.1183
1.9888	0.6170	60	1.8794
1.7142	0.7198	70	1.8562
1.8108	0.8226	80	1.8650
1.7122	0.9254	90	1.8313
1.5926	1.0283	100	1.8228
1.6881	1.1311	110	1.8095
1.4018	1.2339	120	1.8231
1.7688	1.3368	130	1.7767
1.4843	1.4396	140	1.7667
1.5288	1.5424	150	1.7607
1.5946	1.6452	160	1.7498
1.3308	1.7481	170	1.7463
1.6712	1.8509	180	1.7454
1.3342	1.9537	190	1.7449
1.5714	2.0566	200	1.7448

Framework versions

• PEFT 0.13.1.dev0
• Transformers 4.45.1
• Pytorch 2.4.1+cu121
• Datasets 3.0.1
• Tokenizers 0.20.0