_target_: src.models.mllm.seed_x.ContinuousLVLM.from_pretrained
input_resampler: 
  _target_: src.models.tokenizer.qwen_visual.Resampler
  grid_size: 8
  embed_dim: 5120
  num_heads: 32
  kv_dim: 4096

output_resampler:
  _target_: src.models.tokenizer.qwen_visual.Resampler
  grid_size: 8
  embed_dim: 4096
  num_heads: 32
  kv_dim: 5120

add_patch_pos: True 
vit_down: True
mse: True

lm_loss_scale: 1.0
rec_loss_scale: 6.0

pretrained_model_path: https://huggingface.co/AILab-CVC/SEED-X-17B/blob/main/seed_x_i/agent/pytorch_model.bin