jjee2/vincha77__llama381binstruct_summarize_short

llama381binstruct_summarize_short

This model is a fine-tuned version of NousResearch/Meta-Llama-3.1-8B-Instruct on the generator dataset. It achieves the following results on the evaluation set:

• Loss: 2.7140

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: 1
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 30
• training_steps: 500

Training results

Training Loss	Epoch	Step	Validation Loss
1.7457	1.25	25	1.6607
0.7275	2.5	50	1.6843
0.3455	3.75	75	1.7718
0.1936	5.0	100	2.0273
0.108	6.25	125	2.1275
0.0358	7.5	150	2.4614
0.0226	8.75	175	2.3966
0.0184	10.0	200	2.5144
0.0152	11.25	225	2.4407
0.005	12.5	250	2.6186
0.0044	13.75	275	2.6363
0.0053	15.0	300	2.6377
0.0033	16.25	325	2.5876
0.0045	17.5	350	2.6368
0.0022	18.75	375	2.6578
0.0021	20.0	400	2.6785
0.0021	21.25	425	2.6954
0.0018	22.5	450	2.7059
0.002	23.75	475	2.7118
0.0018	25.0	500	2.7140

Framework versions

• PEFT 0.12.0
• Transformers 4.44.2
• Pytorch 2.4.0+cu121
• Datasets 3.0.0
• Tokenizers 0.19.1