Hugging Face's logo

Logolabs
/
agate-preview-002

Text-to-Image
ONNX
Safetensors
English
agate
flow-matching
rectified-flow
compact
256px
logolabs
eurohpc
Model card Files
xet
 Community
agate-preview-002
File size: 2,215 Bytes
49afe28
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
{
  "model_type": "agate",
  "arch": "fcdm_t2",
  "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
    ]
  },
  "params_generator": 190929092,
  "text_encoder": "text_encoder",
  "text_max_len": 512,
  "guide": {
    "generator": "guide/generator.safetensors",
    "text_encoder": "guide/text_encoder",
    "text_max_len": 128,
    "step": 27600
  },
  "vae": "stabilityai/sd-vae-ft-mse",
  "vae_scale": 0.18215,
  "fast_vae": "madebyollin/taesd",
  "resolution": 256,
  "latent_hw": 32,
  "sampler": {
    "kind": "euler",
    "steps": 50,
    "cfg": 3.0,
    "t0_is_noise": true,
    "uncond_prompt": ""
  },
  "training": {
    "step": 138010,
    "samples_seen": 186634240,
    "continued_from": "T40r caption phase, step 82,800 (the run that became Agate Preview 001)",
    "gpu_hours_gh200_lineage": 187.7,
    "energy_kwh_at_wall_lineage": 107.1,
    "co2e_kg_lineage": 3.21,
    "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)"
  },
  "training_compute_flop": {
    "estimate": 6.5e+19,
    "range": [
      5.8e+19,
      7.2e+19
    ],
    "including_text_encoder_pretraining": 8.6e+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": 94.4
    }
  },
  "output_marking": {
    "watermark": "invisible-watermark dwtDctSvd",
    "payload": "AGATE002",
    "metadata_keys": [
      "ai_generated",
      "generator",
      "model",
      "watermark"
    ],
    "default": "on",
    "detect": "agate.detect_watermark(img)"
  }
}