This is an official checkpoint from the paper: "Linear Model Merging Unlocks Simple and Scalable Multimodal Data Mixture Optimization
" (link).
See the official implementation for more information on how to use the models.
qwen2_7b_lora_expert_counting-102400
This model is a fine-tuned version of Qwen/Qwen2-VL-7B on a custom dataset with Counting data (~100k samples).
It achieves the following results on the evaluation set:
Loss: 0.1153
Training hyperparameters
The following hyperparameters were used during training:
learning_rate: 2e-05
train_batch_size: 8
eval_batch_size: 1
seed: 42
distributed_type: multi-GPU
num_devices: 4
gradient_accumulation_steps: 4
total_train_batch_size: 128
total_eval_batch_size: 4
optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
