lora_weights: https://huggingface.co/oshi-pog/p-image-ezsketch-v1/resolve/main/weights.safetensors
hf_api_token:  [a token permitted to read this private repository]
lora_scale:    1.0
prompt:        ezsketch_v1, your subject and scene description

Initial evaluation scale is 1.0. The artifact has not yet been benchmarked across multiple scales; suggested comparison values are 0.75, 1.0, and 1.25 with the same seed and prompt.

Artifact format

weights.safetensors begins with REPLORA1. Despite the extension, it is a Pruna provider container rather than generic Safetensors. It is intended for prunaai/p-image-lora; generic Safetensors and Diffusers loaders will reject it.

Checkpoint: step 1000 (sole checkpoint emitted)
Weight SHA-256: 5192313a31a36a72bb7fcdc6e5740fd991964c42e9e439163d94cb94a5382a7a
Weight bytes: 85092532

Training

text

Trainer:          prunaai/p-image-trainer
Training type:    style
Steps:            1000
Learning rate:    0.0001
Dataset:          100 PNG + 100 matching TXT captions
Default caption:  empty/unset
Trigger:          ezsketch_v1

Status and rights

Status: private, unreviewed pilot
Rights note: private evaluation; audit teacher licence before public or production use
Do not publish or deploy commercially until the teacher provenance and licence are cleared.

See PROVENANCE.json and SHA256SUMS.txt for the local training/output record.