htlou/backup_0202_llamafactory_Compcap_cooccur_0_70-llava-mistral

Compcap_cooccur_0_70

This model is a fine-tuned version of llava-hf/llava-v1.6-mistral-7b-hf on the Compcap_cooccur_0_70 dataset. It achieves the following results on the evaluation set:

• Loss: 0.8001

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-06
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• distributed_type: multi-GPU
• num_devices: 8
• gradient_accumulation_steps: 4
• total_train_batch_size: 256
• total_eval_batch_size: 64
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 10
• num_epochs: 3.0

Training results

Training Loss	Epoch	Step	Validation Loss
0.9216	0.1934	50	0.9377
0.872	0.3868	100	0.8799
0.8452	0.5803	150	0.8544
0.8444	0.7737	200	0.8375
0.8236	0.9671	250	0.8248
0.776	1.1605	300	0.8194
0.7598	1.3540	350	0.8137
0.7539	1.5474	400	0.8075
0.7273	1.7408	450	0.8021
0.7314	1.9342	500	0.7983
0.7094	2.1277	550	0.8029
0.7073	2.3211	600	0.8018
0.6944	2.5145	650	0.8011
0.6841	2.7079	700	0.8003
0.6832	2.9014	750	0.8001

Framework versions

• Transformers 4.45.2
• Pytorch 2.4.0+cu121
• Datasets 2.21.0
• Tokenizers 0.20.3