Rishitha05479/data266-lab2-sd15-lora-run1

DATA 266 Part 2 : LoRA Rank 4

Stable Diffusion 1.5 fine-tuned on the Naruto BLIP Captions dataset for SJSU DATA 266 Lab 2.

Experiment: run1

Selected final model: No

Training

Setting	Value
LoRA rank	4
LoRA alpha	4
Trainable parameters	797,184
Learning rate	0.0001
Steps	4000
Batch size	1
Effective batch size	4
Resolution	512 x 512
Seed	266
GPU hours	2.369
Peak GPU memory	4.38 GB

Evaluation

Metric	Value
Original KID (lower is better)	0.02722
Matched KID (12 generated images)	0.0267
CLIP score	0.2999
LPIPS diversity	0.6941

The matched KID calculation uses identical unblocked in-domain prompt/seed combinations for all three models.

Limitations

The dataset is small and focused on Naruto-style imagery. Fine-tuning improves domain-specific appearance but can reduce prompt adherence, especially for out-of-domain and compositional prompts.

Some generated images were blocked by the safety checker. Blocked outputs were excluded from the image-quality metrics.

The small KID sample size limits the strength of statistical conclusions.

License and intended use

The base model is released under CreativeML OpenRAIL-M. The training images depict copyrighted characters, so this adapter is for non-commercial research and coursework only and must not be used to produce or distribute material commercially.