Text-to-Image
English
numpy
machine-learning
deep-learning
generative-ai
text-2-image
image-generation
open-weights
model-weights
ai-art
pixel-art
game-development
gamedev
game-assets
asset-generator
sprite-generator
offline
tiny-model
numpy-runtime
int8-quantization
self-supervised
procedural-data
gpt
english-prompts
awesome-ai
| { | |
| "model": "PXG-Tiny v0.4", | |
| "params": 483040, | |
| "author": "Chowdhury Tarul Ahsan", | |
| "author_email": "tarulahsan@gmail.com", | |
| "quantization": "all-fp32 precision bundle (no weight quantization) -- guaranteed zero quantization drift for language conditioning and decoding; fp32 arrays retained under the .npz container name for runtime compatibility", | |
| "bundle_bytes": 1809878, | |
| "parity": { | |
| "teacher_forced_max_abs_logit_err": 2e-05, | |
| "argmax_match": 1.0, | |
| "greedy_rollout_agreement": 0.73861 | |
| }, | |
| "val_ce": 0.028889039531350136, | |
| "step": 8200, | |
| "corpus": { | |
| "rows": 97005, | |
| "classes": 38, | |
| "img": "16x16 master-palette indices", | |
| "license": "self-generated (procedural), shipped CC0", | |
| "augmentation": "paraphrase self-teaching, counterfactual twins" | |
| }, | |
| "license": "MIT", | |
| "train_data_hook": "OpenGameArt-CC0 ingestion deferred (procedural ground truth gives exact caption alignment at tiny scale)", | |
| "val_top1": 0.9889424443244934, | |
| "verification": { | |
| "full_grid_38_classes": { | |
| "pass": 152, | |
| "total": 152, | |
| "rate": 1.0, | |
| "note": "2 prompts x 2 seeds per class, user-path pipeline" | |
| }, | |
| "fresh_phrasing_sweep": { | |
| "pass": 116, | |
| "total": 117, | |
| "rate": 0.9915 | |
| }, | |
| "showcase_gallery": { | |
| "pass": 31, | |
| "total": 31 | |
| }, | |
| "raw_heldout_retry_disabled": { | |
| "mean_pass": 0.5942, | |
| "strict_3of3": 0.5, | |
| "n_prompts": 46 | |
| }, | |
| "fixes": [ | |
| "head-noun parser (wizard staff)", | |
| "fighter-word fallback", | |
| "plural class nouns", | |
| "idiom-safe glow (shining armor)", | |
| "capability-filtered material/attribute checks", | |
| "GT-calibrated leg/wing/extent thresholds", | |
| "escalating self-guidance + spatial priors + canonical anchor", | |
| "deer structural antler prefix (AR cascade fix)" | |
| ] | |
| } | |
| } |