bryan111232/OmniNWM-VLA

stage2_llama_factory

This model is a fine-tuned version of /code/VLA/models/Qwen2.5-VL-7B-Instruct on the tri_modal_fused_train dataset. It achieves the following results on the evaluation set:

• Loss: 0.0843

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: 2
• eval_batch_size: 4
• seed: 42
• distributed_type: multi-GPU
• num_devices: 8
• gradient_accumulation_steps: 8
• total_train_batch_size: 128
• total_eval_batch_size: 32
• 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: 8

Training results

Training Loss	Epoch	Step	Validation Loss
0.0969	1.3889	250	0.0951
0.0869	2.7778	500	0.0876
0.0868	4.1667	750	0.0856
0.0842	5.5556	1000	0.0848
0.0821	6.9444	1250	0.0843

Framework versions

• PEFT 0.14.0
• Transformers 4.49.0
• Pytorch 2.2.0+cu121
• Datasets 2.16.0
• Tokenizers 0.21.1