omnisvg-m1-finetune-plain-19bbf5b-32icons-9b9b1699
completed —
Measured result
Verbatim from the run's summary.json.
checks
clip_gain_over_control
false
ended_rate
false
reference_top1_rate
false
clip
repo
openai/clip-vit-base-patch32
revision
3d74acf9a28c67741b2f4f2ea7635f0aaf6f0268
clip_to_caption
mean
0.21794391772709787
median
0.21727204322814941
n
64
clip_to_reference
mean
0.7724125878885388
median
0.7812369465827942
n
64
codec_valid_rate
0.078125
config_sha256
f3b8c600499b331a…control
clip_to_reference_mean
0.7724125878885388
control_clip_to_reference_mean
0.8112438190728426
control_codec_valid_rate
0.734375
control_reference_top1_rate
0.078125
control_rows_sha256
3f6131b329d040a5…gain_ci95
-0.055812961468473074, -0.022486942750401796
icons_improved
8
mean_gain
-0.03883123118430376
paired_icons
32
report_root
reports/learning/omnisvg-m1-control
criteria
clip_gain_over_control_ci_excludes_zero
true
min_ended_rate
0.9
min_reference_top1_rate
0.25
decoded_rate
1.0
drawings
64
ended_rate
0.453125
failures
encode:control coordinate is outside -8..96 and clamping is disabled
2
encode:endpoint coordinate is outside 0..72 and clamping is disabled
10
encode:program needs 10 contours and up to 257 segments
1
encode:program needs 10 contours and up to 311 segments
1
encode:program needs 10 contours and up to 354 segments
1
encode:program needs 10 contours and up to 395 segments
1
encode:program needs 10 contours and up to 473 segments
1
encode:program needs 10 contours and up to 507 segments
1
encode:program needs 10 contours and up to 522 segments
1
encode:program needs 11 contours and up to 161 segments
1
encode:program needs 11 contours and up to 378 segments
1
encode:program needs 11 contours and up to 405 segments
1
encode:program needs 13 contours and up to 145 segments
1
encode:program needs 13 contours and up to 310 segments
1
encode:program needs 13 contours and up to 379 segments
1
encode:program needs 13 contours and up to 437 segments
1
encode:program needs 15 contours and up to 139 segments
1
encode:program needs 2 contours and up to 516 segments
1
encode:program needs 2 contours and up to 696 segments
1
encode:program needs 3 contours and up to 146 segments
1
encode:program needs 3 contours and up to 147 segments
1
encode:program needs 3 contours and up to 208 segments
1
encode:program needs 4 contours and up to 157 segments
1
encode:program needs 4 contours and up to 487 segments
1
encode:program needs 4 contours and up to 495 segments
1
encode:program needs 4 contours and up to 536 segments
1
encode:program needs 4 contours and up to 556 segments
1
encode:program needs 4 contours and up to 703 segments
1
encode:program needs 5 contours and up to 216 segments
1
encode:program needs 5 contours and up to 305 segments
1
encode:program needs 5 contours and up to 509 segments
1
encode:program needs 5 contours and up to 551 segments
1
encode:program needs 6 contours and up to 248 segments
1
encode:program needs 6 contours and up to 277 segments
1
encode:program needs 6 contours and up to 333 segments
1
encode:program needs 6 contours and up to 350 segments
1
encode:program needs 6 contours and up to 489 segments
1
encode:program needs 6 contours and up to 502 segments
1
encode:program needs 7 contours and up to 227 segments
1
encode:program needs 7 contours and up to 541 segments
1
encode:program needs 8 contours and up to 320 segments
1
encode:program needs 8 contours and up to 432 segments
1
encode:program needs 8 contours and up to 492 segments
1
encode:program needs 9 contours and up to 270 segments
1
encode:program needs 9 contours and up to 345 segments
1
pack:program needs 136 packed segments but capacity is 128
2
pack:program needs 206 packed segments but capacity is 128
1
pack:program needs 213 packed segments but capacity is 128
1
icons
32
max_new_tokens
2048
mean_snap_distance_rgb
6.419055435476086
median_paths
5.5
median_tokens
2048.0
model
adapter
data/processed/prior/omnisvg-m1-finetune/adapter.zip
adapter_sha256
23bf01ba394a1e04…checkpoint_sha256
39f36759fc677b32…fine_tuned
true
parameters
3846914048
repo
OmniSVG/OmniSVG1.1_4B
revision
117d4c4541839e02943554fe71c0a0d54fcef819
predeclared_outcome
falsified
prompt_style
training
reference_rank
chance_top1
0.03125
mean
15.953125
scored
64
top1_rate
0.046875
top5_rate
0.171875
rows_sha256
c931085429378016…samples_per_icon
2
sampling
repetition_penalty
1.0
temperature
1.0
top_k
0
top_p
0.95
schema_version
1
sheet_sha256
cca4acf24eac60ea…study_version
omnisvg-m1-finetune-plain
timing
generate_seconds
1207.0101925859635
load_seconds
37.807317397004226
peak_vram_gib
8.129339694976807
seconds_per_drawing
18.85953425915568
Visual output
State transitions
- planned2026-09-21T14:46:57Z
- completed2026-09-21T15:21:20Zat temperature 1.0 with no repetition penalty the adapter draws black scribbles: 55% of drawings run to the 2,048-token cap, CLIP-to-reference 0.772 (below the control's 0.811, paired gain -0.039 with an interval of -0.056 to -0.022), the right icon first for 3 of 64
Run record
This run has no run.yaml. What follows is the identity and configuration carried by its rows in state/runs.jsonl, the append-only registry.
run_id
omnisvg-m1-finetune-plain-19bbf5b-32icons-9b9b1699
config
configs/learning/omnisvg-m1-finetune-plain.yaml
outputs
report_root
reports/learning/omnisvg-m1-finetune-plain