Add correction-jit-bs1024-h200-20261010 latest checkpoint config
Browse files
correction-training-private/correction-jit-bs1024-h200-20261010/step-0100000/config.json
ADDED
|
@@ -0,0 +1,81 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"format": "correction-training/2",
|
| 3 |
+
"kind": "pair",
|
| 4 |
+
"model": {
|
| 5 |
+
"model": "JiT-L/16",
|
| 6 |
+
"image_size": 256,
|
| 7 |
+
"num_classes": 1000,
|
| 8 |
+
"t_eps": 0.05,
|
| 9 |
+
"noise_scale": 1.0,
|
| 10 |
+
"p_mean": -0.8,
|
| 11 |
+
"p_std": 0.8,
|
| 12 |
+
"label_drop": 0.1,
|
| 13 |
+
"prediction": "x",
|
| 14 |
+
"data_time": 1,
|
| 15 |
+
"time_sampling": "logit_normal"
|
| 16 |
+
},
|
| 17 |
+
"codec": {
|
| 18 |
+
"name": "identity",
|
| 19 |
+
"pretrained": "stabilityai/sd-vae-ft-ema",
|
| 20 |
+
"revision": null,
|
| 21 |
+
"scale": 0.18215,
|
| 22 |
+
"bias": 0.0
|
| 23 |
+
},
|
| 24 |
+
"correction": {
|
| 25 |
+
"input": "x",
|
| 26 |
+
"output": "x",
|
| 27 |
+
"combine": "direct",
|
| 28 |
+
"gate_init": 0.0,
|
| 29 |
+
"gate_shape": "channel",
|
| 30 |
+
"gate_mix": "difference"
|
| 31 |
+
},
|
| 32 |
+
"training": {
|
| 33 |
+
"loss_space": "v",
|
| 34 |
+
"detach_input": true,
|
| 35 |
+
"detach_skip": true,
|
| 36 |
+
"deterministic": false,
|
| 37 |
+
"correction_steps": 1,
|
| 38 |
+
"base_loss_weight": 1.0,
|
| 39 |
+
"correction_loss_weight": 1.0,
|
| 40 |
+
"steps": 100000,
|
| 41 |
+
"batch_size": 64,
|
| 42 |
+
"accumulation": 2,
|
| 43 |
+
"lr": 5e-05,
|
| 44 |
+
"betas": [
|
| 45 |
+
0.9,
|
| 46 |
+
0.95
|
| 47 |
+
],
|
| 48 |
+
"ema": [
|
| 49 |
+
0.9999,
|
| 50 |
+
0.9996
|
| 51 |
+
],
|
| 52 |
+
"precision": "bf16",
|
| 53 |
+
"gradient_precision": "fp32",
|
| 54 |
+
"gradient_bucket_mb": 25,
|
| 55 |
+
"warmup_epochs": 5.0,
|
| 56 |
+
"save_every": 1000,
|
| 57 |
+
"export_every": 5000
|
| 58 |
+
},
|
| 59 |
+
"sampling": {
|
| 60 |
+
"method": "heun",
|
| 61 |
+
"precision": "bf16",
|
| 62 |
+
"steps": 45,
|
| 63 |
+
"corrections": 9,
|
| 64 |
+
"correction_interval": [
|
| 65 |
+
0.8,
|
| 66 |
+
1.0
|
| 67 |
+
],
|
| 68 |
+
"cfg": 1.0,
|
| 69 |
+
"cfg_interval": [
|
| 70 |
+
0.1,
|
| 71 |
+
1.0
|
| 72 |
+
]
|
| 73 |
+
},
|
| 74 |
+
"source": {
|
| 75 |
+
"initialization": "eba12888010d531c4e0dcc75bdf30dbdcf7eeb7a812affc52fa5e3b8b0e183ae",
|
| 76 |
+
"step": 100000,
|
| 77 |
+
"ema": "model_ema2",
|
| 78 |
+
"dataset": "imagenette2-320/folder-v1"
|
| 79 |
+
},
|
| 80 |
+
"weights_sha256": "4d044ed9c450aaa0153ea47073dec1c0b09722b30f642429760953e383740c2d"
|
| 81 |
+
}
|