Hugging Face's logo

Logolabs
/
agate-preview-003

Text-to-Image
ONNX
Safetensors
English
agate
flow-matching
rectified-flow
compact
512px
logolabs
eurohpc
Model card Files
xet
 Community
1
agate-preview-003
File size: 3,019 Bytes
4770b6a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
{
  "model_type": "agate",
  "arch": "fcdm_t2mr",
  "model_kw": {
    "widths": [
      288,
      832,
      1664
    ],
    "depths": [
      4,
      6,
      2
    ],
    "steer_rank": [
      128,
      256,
      256
    ],
    "ctx_dim": 512,
    "dim": 640,
    "prelude": 3,
    "core": 3,
    "coda": 2,
    "loops": 4,
    "loop_probs": [
      0.0,
      0.0,
      0.0,
      1.0
    ],
    "read_ch": 96,
    "outer_width": 192,
    "outer_depth": [
      2,
      2
    ],
    "up_rank": 32,
    "outer_mode": "flat",
    "hp_prior": true,
    "coord_res": true
  },
  "params_generator": 194270952,
  "params_generator_active": 192148612,
  "text_encoder": "text_encoder",
  "text_max_len": 512,
  "prompt_pipeline": {
    "normalize": true,
    "spell": true,
    "negatives_from_prompt": true,
    "count_code": true,
    "spell_separator": " || spell: "
  },
  "vae": "stabilityai/sd-vae-ft-mse",
  "vae_scale": 0.18215,
  "fast_vae": "madebyollin/taesd",
  "resolution": 512,
  "resolutions": {
    "512": {
      "latent_hw": 64,
      "shift": 2.0
    },
    "256": {
      "latent_hw": 32,
      "shift": 1.0
    }
  },
  "sampler": {
    "kind": "euler",
    "steps": 50,
    "cfg": 3.0,
    "t0_is_noise": true,
    "uncond_prompt": "",
    "timestep_shift": "SD3 shift_t, 2.0 at 512 px, 1.0 at 256 px"
  },
  "training": {
    "step": 7400,
    "samples_seen": 196597760,
    "lineage": [
      "T40r / cap2 continuation, step 121,440 (pre-anneal, 256 px)",
      "mr512_long1: 12,000 steps, batch 512, 256+512 px (75% 512), T40r weights at 0.5x lr",
      "mr512_long2: 5,500 steps, batch 1,024, everything trainable incl. Ettin at 5e-5",
      "mr512_perc1: 7,400 steps, batch 2,048, saliency-weighted flow + 0.1 LPIPS + 0.01 P-DINO, Ettin frozen"
    ],
    "gpu_hours_gh200_lineage": 282.8,
    "energy_kwh_at_wall_lineage": 164.5,
    "co2e_kg_lineage": 4.94,
    "dataset": "LucasFang/FLUX-Reason-6M",
    "lineage_note": "whole lineage from scratch, jobs pro-rated by steps where this release branched mid-run (perun ledger, PUE 1.3, 30 g/kWh)",
    "data_prep_gpu_hours_gh200": 22.5
  },
  "training_compute_flop": {
    "estimate": 7.8e+19,
    "range": [
      7.1e+19,
      8.5e+19
    ],
    "including_text_encoder_pretraining": 8.8e+20,
    "method": "measured forward FLOPs per image (torch.utils.flop_counter, batch 1, 128-token context) x images seen x 3 (forward + backward); Ettin at 128 tokens (low) to 4x (high); excludes the compute used by third parties to generate the synthetic training images",
    "forward_gflop_per_image": {
      "256": 97.5,
      "512": 255.9
    }
  },
  "output_marking": {
    "watermark": "invisible-watermark dwtDctSvd",
    "payload": "AGATE003",
    "metadata_keys": [
      "ai_generated",
      "generator",
      "model",
      "watermark"
    ],
    "default": "on",
    "detect": "agate.detect_watermark(img)"
  }
}