old_13M/config.json

{
  "_name_or_path": "/home/ahf38/palmer_scratch/brainlm/training-runs/2023-07-17-19_00_00/checkpoint-1750",
  "architectures": [
    "BrainLMForPretraining"
  ],
  "attention_probs_dropout_prob": 0.1,
  "conv_kernel_size": 65,
  "decoder_hidden_size": 512,
  "decoder_intermediate_size": 1024,
  "decoder_num_attention_heads": 4,
  "decoder_num_hidden_layers": 2,
  "hidden_act": "gelu",
  "hidden_dropout_prob": 0.0,
  "hidden_size": 512,
  "image_size": 224,
  "initializer_range": 0.02,
  "intermediate_size": 1024,
  "inv_coeff_init_option": false,
  "layer_norm_eps": 1e-12,
  "loss_fn": "mse",
  "mask_ratio": 0.2,
  "max_eval_samples": 800,
  "model_type": "brainlm_mae",
  "norm_pix_loss": false,
  "num_attention_heads": 4,
  "num_brain_voxels": 424,
  "num_channels": 3,
  "num_hidden_layers": 4,
  "num_landmarks": 64,
  "num_timepoints_per_voxel": 200,
  "patch_size": 16,
  "qkv_bias": true,
  "segment_means_seq_len": 64,
  "timepoint_patching_size": 20,
  "torch_dtype": "float32",
  "transformers_version": "4.30.0.dev0",
  "use_tanh_decoder": false,
  "weight_decay": 1e-05
}