Upload /linear-sit-img256-bs512/args.json with huggingface_hub
Browse files
linear-sit-img256-bs512/args.json
ADDED
|
@@ -0,0 +1,45 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"output_dir": "exps",
|
| 3 |
+
"exp_name": "linear-sssit-bn-adaln-proj0.5-triu0.025-enc8-img256-bs256-trand",
|
| 4 |
+
"logging_dir": "logs",
|
| 5 |
+
"report_to": "wandb",
|
| 6 |
+
"sampling_steps": 10000,
|
| 7 |
+
"resume_step": 300000,
|
| 8 |
+
"model": "SiT-XL/2",
|
| 9 |
+
"num_classes": 1000,
|
| 10 |
+
"encoder_depth": 8,
|
| 11 |
+
"fused_attn": true,
|
| 12 |
+
"qk_norm": false,
|
| 13 |
+
"data_dir": "/mnt/main_storage/peihao/diffusion/datasets/imagenet_256",
|
| 14 |
+
"resolution": 256,
|
| 15 |
+
"batch_size": 256,
|
| 16 |
+
"allow_tf32": true,
|
| 17 |
+
"mixed_precision": "fp16",
|
| 18 |
+
"epochs": 1400,
|
| 19 |
+
"max_train_steps": 400000,
|
| 20 |
+
"checkpointing_steps": 50000,
|
| 21 |
+
"gradient_accumulation_steps": 1,
|
| 22 |
+
"learning_rate": 0.0001,
|
| 23 |
+
"adam_beta1": 0.9,
|
| 24 |
+
"adam_beta2": 0.999,
|
| 25 |
+
"adam_weight_decay": 0.0,
|
| 26 |
+
"adam_epsilon": 1e-08,
|
| 27 |
+
"max_grad_norm": 1.0,
|
| 28 |
+
"seed": 0,
|
| 29 |
+
"num_workers": 4,
|
| 30 |
+
"path_type": "linear",
|
| 31 |
+
"prediction": "v",
|
| 32 |
+
"cfg_prob": 0.1,
|
| 33 |
+
"proj_dims": [
|
| 34 |
+
2048,
|
| 35 |
+
768
|
| 36 |
+
],
|
| 37 |
+
"proj_adaln": true,
|
| 38 |
+
"proj_bn": true,
|
| 39 |
+
"proj_coeff": 0.5,
|
| 40 |
+
"triu_coeff": 0.025,
|
| 41 |
+
"proj_global_batch": true,
|
| 42 |
+
"rand_t_batch": true,
|
| 43 |
+
"weighting": "uniform",
|
| 44 |
+
"legacy": false
|
| 45 |
+
}
|