htlou/mm-interp-Align-Anything-Coccur-q0_25

Align-Anything-Coccur-q0_25

This model is a fine-tuned version of llava-hf/llava-v1.6-mistral-7b-hf on the Align-Anything-Coccur-q0_25 dataset. It achieves the following results on the evaluation set:

• Loss: 0.8606

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
0.9472	0.5731	50	0.9285
0.7918	1.1461	100	0.8879
0.7632	1.7192	150	0.8642
0.6847	2.2923	200	0.8615
0.685	2.8653	250	0.8606

Framework versions

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