llama_2_o1_10_full

This model is a fine-tuned version of meta-llama/Llama-2-7b-chat-hf on the generator dataset. It achieves the following results on the evaluation set:

• Loss: 0.6120

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: 2e-05
• train_batch_size: 4
• eval_batch_size: 4
• seed: 42
• distributed_type: multi-GPU
• num_devices: 4
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• total_eval_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 1

Training results

Training Loss	Epoch	Step	Validation Loss
0.9433	0.0615	100	0.7648
0.7768	0.1230	200	0.7311
0.7448	0.1844	300	0.7125
0.7336	0.2459	400	0.6952
0.7224	0.3074	500	0.6859
0.714	0.3689	600	0.6708
0.6965	0.4304	700	0.6594
0.6883	0.4919	800	0.6585
0.6738	0.5533	900	0.6418
0.6712	0.6148	1000	0.6328
0.6633	0.6763	1100	0.6239
0.6641	0.7378	1200	0.6183
0.6578	0.7993	1300	0.6167
0.6519	0.8607	1400	0.6141
0.652	0.9222	1500	0.6122
0.655	0.9837	1600	0.6117

Framework versions

• Transformers 4.44.2
• Pytorch 2.4.1+cu121
• Datasets 3.0.0
• Tokenizers 0.19.1