p-image-trainer

p-image-trainer allows you to train custom LoRA (Low-Rank Adaptation) weights for use with the p-image-lora model. Train personalized styles, subjects, characters, or concepts using your own images.

This is NOT an inference model. It does not generate images. Instead, it outputs a ZIP file containing trained LoRA weights (.safetensors).

Rate Limit: 5 requests per minute

Category: LoRA Training

Price: $1.80 / 1000 steps

Important Notes:

• Async only - Training takes minutes to hours. Do not use Try-Sync header.
• Download within 30 minutes - The output URL expires approximately 30 minutes after training completes. Download immediately and upload to HuggingFace for permanent storage.
• Trained LoRAs only work with p-image-lora, not with other models.

Workflow

1. Prepare training data - Create a ZIP file with at least 10 images and optional caption files
2. Upload ZIP to accessible URL - Host your training data somewhere accessible (e.g., cloud storage)
3. Start training - Submit async request (takes minutes to hours)
4. Poll for completion - Check status until training succeeds
5. Download output - Get the ZIP file within 30 minutes
6. Upload to HuggingFace - Store the .safetensors file for permanent access
7. Use with p-image-lora - Generate images using your trained LoRA

Quickstart

Prepare Training Data

Create a ZIP archive with your training images:

training_data.zip
├── image1.jpg
├── image1.txt        # Optional: caption for image1.jpg
├── image2.png
├── image2.txt        # Optional: caption for image2.png
├── image3.jpg
└── ...

Use at least 10 images for good results. Caption files should match image names (e.g., photo.txt for photo.jpg).

Start Training (Async Only)

curl -X POST 'https://api.pruna.ai/v1/predictions' \
  -H 'Content-Type: application/json' \
  -H 'apikey: YOUR_API_KEY' \
  -H 'Model: p-image-trainer' \
  -d '{
    "input": {
      "image_data": "https://your-storage.com/training_data.zip",
      "steps": 1000,
      "training_type": "balanced"
    }
  }'

Response:

{
  "id": "training123xyz",
  "model": "p-image-trainer",
  "input": { ... },
  "get_url": "https://api.pruna.ai/v1/predictions/status/training123xyz"
}

Poll for Completion

Training takes minutes to hours depending on steps. Poll periodically:

curl -X GET 'https://api.pruna.ai/v1/predictions/status/training123xyz' \
  -H 'apikey: YOUR_API_KEY'

When complete:

{
  "status": "succeeded",
  "output": "https://api.pruna.ai/v1/predictions/delivery/xezq/abc123.../lora_weights.zip"
}

Download Output Immediately

The output URL expires in ~30 minutes. Download the ZIP file immediately:

curl -o lora_output.zip "https://api.pruna.ai/v1/predictions/delivery/xezq/abc123.../lora_weights.zip"

Upload to HuggingFace

Extract and upload the .safetensors file to HuggingFace:

unzip lora_output.zip
# Upload lora.safetensors to huggingface.co/your-username/my-lora

Use with p-image-lora

curl -X POST 'https://api.pruna.ai/v1/predictions' \
  -H 'Content-Type: application/json' \
  -H 'apikey: YOUR_API_KEY' \
  -H 'Model: p-image-lora' \
  -d '{
    "input": {
      "prompt": "A portrait in my custom style",
      "lora_weights": "huggingface.co/your-username/my-lora"
    }
  }'

Parameters

Required Parameters

Parameter	Type	Description
image_data	string (URI)	URL to a ZIP archive with training images. Use at least 10 images. Can include text files for captions (e.g., photo.txt for photo.jpg)

Optional Parameters

Parameter	Type	Default	Description
steps	integer	1000	Number of training steps. Range: 100-5000, in increments of 100. More steps = longer training, potentially better results
learning_rate	number	0.0001	Learning rate for training. Range: 0.00001-0.01. Lower = slower but more stable
training_type	string	"balanced"	Type of training: "content" (subjects/characters), "style" (artistic styles), "balanced" (both)
default_caption	string	-	Default caption for images without .txt files. If not provided and captions are missing, training will fail

Training Type Options

Type	Use Case
content	Specific subjects, characters, objects, people
style	Artistic styles, color palettes, aesthetic treatments
balanced	General-purpose, mixed content and style

Steps Guidelines

Steps	Use Case	Expected Time
100-500	Quick tests, simple styles	Minutes
500-1000	Standard training	10-30 minutes
1000-2000	High quality, complex	30-60 minutes
2000-5000	Maximum quality	1-2+ hours