Upload 256T_NC=262144.yaml

256T_NC=262144.yaml

@@ -0,0 +1,42 @@
+workspace: /path/to/workspace
+name: 256T_NC=262144
+log_path: ${workspace}/outputs/${name}
+world_size: 1
+seed: 42
+
+data:
+  train_data_path: /path/to/converted_imagenet_train
+  val_data_path: /path/to/converted_imagenet_val
+  batch_size: 128
+  world_size: ${world_size}
+
+model:
+  vae_path: ${workspace}/ckpt/dc_ae
+  log_path: ${log_path}
+  input_data_size: [32, 8, 8]
+  quantizer:
+    TYPE: vq
+    num_code: 262144
+    num_group: 256
+    tokens_per_data: 256
+  decoder:
+    TYPE: dc_ae
+    in_channels: 32
+    latent_channels: 32
+    attention_head_dim: 32
+    block_type: [EfficientViTBlock]
+    block_out_channels: [512]
+    layers_per_block: [4]
+    qkv_multiscales: [[5]]
+    norm_type: rms_norm
+    act_fn: silu
+    upsample_block_type: interpolate
+
+train:
+  log_path: ${log_path}
+  batch_size: ${data.batch_size}
+  world_size: ${world_size}
+  num_epochs: 100
+  lr: 0.01
+  visualize_interval: 1
+  image_log_path: ${log_path}/images