all_tasks_combined_8b_sft_more_epochs

This model is a fine-tuned version of meta-llama/Meta-Llama-3-8B-Instruct on the identity and the data_mc_filtered datasets. It achieves the following results on the evaluation set:

• Loss: 0.8986

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: 1e-05
• train_batch_size: 2
• eval_batch_size: 2
• seed: 42
• distributed_type: multi-GPU
• num_devices: 8
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• total_eval_batch_size: 16
• 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: 0.1
• num_epochs: 6.0

Training results

Training Loss	Epoch	Step	Validation Loss
0.4512	0.3714	200	0.5138
0.5062	0.7428	400	0.5233
0.3444	1.1133	600	0.4961
0.3574	1.4847	800	0.4851
0.2927	1.8561	1000	0.4776
0.2063	2.2266	1200	0.5153
0.1942	2.5980	1400	0.5041
0.1876	2.9694	1600	0.4744
0.1046	3.3398	1800	0.5740
0.0851	3.7112	2000	0.5829
0.0381	4.0817	2200	0.7345
0.0402	4.4531	2400	0.6936
0.0295	4.8245	2600	0.7317
0.0105	5.1950	2800	0.8839
0.0082	5.5664	3000	0.8951
0.0092	5.9378	3200	0.8989

Framework versions

• Transformers 4.49.0
• Pytorch 2.5.1+cu124
• Datasets 3.2.0
• Tokenizers 0.21.0