You're looking at a specific version of this model. Jump to the model overview.
lightweight-ai /q_l_t:9aff86c5
Input schema
The fields you can use to run this model with an API. If you don’t give a value for a field its default value will be used.
| Field | Type | Default value | Description |
|---|---|---|---|
| seed |
integer
|
42
|
Random seed
|
| steps |
integer
|
1000
Min: 1 |
Total training steps (optimizer updates)
|
| lora_r |
integer
|
16
Min: 1 |
LoRA rank r
|
| data_zip |
string
|
ZIP of training data. Each image must have a caption file with the same base name and a .txt extension (e.g., cat.jpg + cat.txt). Files can be in subfolders.
|
|
| base_model |
string
|
Qwen/Qwen-Image
|
Base VLM to fine-tune (HuggingFace ID). Use a Qwen-VL/Qwen2-VL instruct checkpoint, e.g. 'Qwen/Qwen2-VL-2B-Instruct'.
|
| batch_size |
integer
|
1
Min: 1 |
Per-device batch size
|
| lora_alpha |
integer
|
32
Min: 1 |
LoRA alpha
|
| resolution |
integer
|
672
Min: 64 |
Short-side image resolution for training (if your job type uses it)
|
| save_every |
integer
|
500
Min: 1 |
Save checkpoint every N steps
|
| resume_from |
string
|
Optional path or HF repo to resume from a previous LoRA checkpoint
|
|
| lora_dropout |
number
|
0.05
Max: 1 |
LoRA dropout
|
| learning_rate |
number
|
0.00005
|
Learning rate
|
| gradient_accumulation_steps |
integer
|
4
Min: 1 |
Gradient accumulation to emulate larger batch sizes
|
Output schema
The shape of the response you’ll get when you run this model with an API.
Schema
{'format': 'uri', 'title': 'Output', 'type': 'string'}