Upload VAR/tokenizer/checkpoints artifacts

VAR/tokenizer/checkpoints/epoch_040_tokenizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9e1aefa735b638ba83db20d144c792db308d3a296ca468c0c1b9c9547ea8f9b0
+size 436084970

VAR/tokenizer/checkpoints/resolved_config.json

@@ -0,0 +1,109 @@
+{
+  "command": [
+    "/mnt/research-common-h100-data/home/dylanlim/repos/MsTok/src/image_tokenization/training/train_var_tokenizer.py",
+    "--config",
+    "configs/baselines/var/tokenizer_imagenet256.yaml"
+  ],
+  "config": {
+    "adam_betas": [
+      0.5,
+      0.9
+    ],
+    "adaptive_epsilon": 0.0001,
+    "adaptive_max": 10000.0,
+    "adversarial_weight": 0.5,
+    "auto_resume": true,
+    "base_channels": 160,
+    "beta": 0.25,
+    "cache_dir": "/mnt/research-common-h100-data/home/dylanlim/repos/MsTok/data/cache",
+    "data_root": "/mnt/research-common-h100-data/home/dylanlim/datasets/imagenet",
+    "epochs": 100,
+    "eval_batch_size": 16,
+    "eval_checkpoint": "",
+    "eval_num_workers": 8,
+    "eval_output": "",
+    "expected_classes": 1000,
+    "expected_parameters": 108948355,
+    "expected_train_images": 1281167,
+    "expected_val_images": 50000,
+    "expected_world_size": 16,
+    "gan_start_epoch": 3,
+    "gradient_clip": 1.0,
+    "grid_epochs": [
+      2,
+      5,
+      10,
+      20,
+      30,
+      40,
+      50,
+      60,
+      70,
+      80,
+      90,
+      100
+    ],
+    "image_size": 256,
+    "latent_channels": 32,
+    "learning_rate": 0.0001,
+    "local_batch_size": 16,
+    "log_interval": 50,
+    "master_port": 29500,
+    "num_workers": 8,
+    "output_dir": "/mnt/research-common-h100-data/home/dylanlim/repos/MsTok/outputs",
+    "patch_nums": [
+      1,
+      2,
+      3,
+      4,
+      5,
+      6,
+      8,
+      10,
+      13,
+      16
+    ],
+    "perceptual_weight": 1.0,
+    "permanent_epochs": [
+      40,
+      60,
+      80,
+      100
+    ],
+    "precision": "bf16",
+    "primary_epoch": 100,
+    "quant_resi": 0.5,
+    "real_stats_cache": "/mnt/research-common-h100-data/home/dylanlim/repos/MsTok/data/cache/imagenet256_val_pytorch_fid_stats.npz",
+    "reconstruction_weight": 1.0,
+    "recovery_interval": 5000,
+    "resize_short": 288,
+    "resume": "",
+    "run_name": "var_imagenet256_v1",
+    "seed": 42,
+    "share_quant_resi": 4,
+    "stop_after_step": 0,
+    "usage_interval": 500,
+    "vocab_size": 4096,
+    "vq_weight": 1.0,
+    "wandb_enabled": true,
+    "wandb_entity": "mstok",
+    "wandb_group": "VAR-tokenizer",
+    "wandb_project": "baselines",
+    "weight_decay": 0.0
+  },
+  "environment": {
+    "cuda_runtime": "12.1",
+    "cudnn": 8902,
+    "lpips": "0.1.4",
+    "numpy": "1.26.4",
+    "pillow": "10.2.0",
+    "python": "3.11.15 (main, Jun 11 2026, 15:20:16) [GCC 14.3.0]",
+    "pytorch_fid": "0.3.0",
+    "scipy": "1.11.4",
+    "torch": "2.1.2+cu121",
+    "torchvision": "0.16.2",
+    "wandb": "0.17.9",
+    "world_size": 16
+  },
+  "git_revision": "961d40ec6dd4023af6605977b76c965a4681d13e"
+}

VAR/tokenizer/checkpoints/tokenizer_imagenet256.yaml

@@ -0,0 +1,61 @@
+run_name: var_imagenet256_v1
+output_dir: $MSTOK_OUTPUT_DIR
+data_root: $IMAGENET_ROOT
+cache_dir: $MSTOK_CACHE_DIR
+
+image_size: 256
+resize_short: 288
+patch_nums: [1, 2, 3, 4, 5, 6, 8, 10, 13, 16]
+vocab_size: 4096
+latent_channels: 32
+base_channels: 160
+beta: 0.25
+quant_resi: 0.5
+share_quant_resi: 4
+expected_parameters: 108948355
+
+epochs: 100
+local_batch_size: 16
+expected_world_size: 16
+num_workers: 8
+seed: 42
+
+learning_rate: 0.0001
+adam_betas: [0.5, 0.9]
+weight_decay: 0.0
+gradient_clip: 1.0
+
+reconstruction_weight: 1.0
+vq_weight: 1.0
+perceptual_weight: 1.0
+adversarial_weight: 0.5
+gan_start_epoch: 3
+adaptive_epsilon: 0.0001
+adaptive_max: 10000.0
+
+precision: bf16
+recovery_interval: 5000
+permanent_epochs: [40, 60, 80, 100]
+primary_epoch: 100
+resume: ""
+auto_resume: true
+stop_after_step: 0
+
+log_interval: 50
+usage_interval: 500
+grid_epochs: [2, 5, 10, 20, 30, 40, 50, 60, 70, 80, 90, 100]
+wandb_enabled: true
+wandb_entity: mstok
+wandb_project: baselines
+wandb_group: VAR-tokenizer
+
+eval_batch_size: 16
+eval_num_workers: 8
+eval_checkpoint: ""
+eval_output: ""
+real_stats_cache: $MSTOK_CACHE_DIR/imagenet256_val_pytorch_fid_stats.npz
+
+expected_train_images: 1281167
+expected_val_images: 50000
+expected_classes: 1000
+master_port: 29500