Instructions to use sprok-daniel-mozaik/LightOnOCR-2-1B-Math-Handwritten with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use sprok-daniel-mozaik/LightOnOCR-2-1B-Math-Handwritten with PEFT:

from peft import PeftModel
from transformers import AutoModelForCausalLM

base_model = AutoModelForCausalLM.from_pretrained("lightonai/LightOnOCR-2-1B")
model = PeftModel.from_pretrained(base_model, "sprok-daniel-mozaik/LightOnOCR-2-1B-Math-Handwritten")

Transformers

How to use sprok-daniel-mozaik/LightOnOCR-2-1B-Math-Handwritten with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="sprok-daniel-mozaik/LightOnOCR-2-1B-Math-Handwritten")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("sprok-daniel-mozaik/LightOnOCR-2-1B-Math-Handwritten", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use sprok-daniel-mozaik/LightOnOCR-2-1B-Math-Handwritten with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "sprok-daniel-mozaik/LightOnOCR-2-1B-Math-Handwritten"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "sprok-daniel-mozaik/LightOnOCR-2-1B-Math-Handwritten",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/sprok-daniel-mozaik/LightOnOCR-2-1B-Math-Handwritten

SGLang

How to use sprok-daniel-mozaik/LightOnOCR-2-1B-Math-Handwritten with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "sprok-daniel-mozaik/LightOnOCR-2-1B-Math-Handwritten" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "sprok-daniel-mozaik/LightOnOCR-2-1B-Math-Handwritten",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "sprok-daniel-mozaik/LightOnOCR-2-1B-Math-Handwritten" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "sprok-daniel-mozaik/LightOnOCR-2-1B-Math-Handwritten",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use sprok-daniel-mozaik/LightOnOCR-2-1B-Math-Handwritten with Docker Model Runner:
```
docker model run hf.co/sprok-daniel-mozaik/LightOnOCR-2-1B-Math-Handwritten
```

LightOnOCR-2-1B-Math-Handwritten

This model is a fine-tuned version of lightonai/LightOnOCR-2-1B on an unknown dataset. It achieves the following results on the evaluation set:

Loss: 0.3156

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 6e-05
train_batch_size: 4
eval_batch_size: 6
seed: 42
gradient_accumulation_steps: 4
total_train_batch_size: 16
optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 10
num_epochs: 1

Training results

Training Loss	Epoch	Step	Validation Loss
1.1458	0.16	50	0.5339
0.4190	0.32	100	0.3892
0.3376	0.48	150	0.3491
0.3011	0.64	200	0.3280
0.2831	0.8	250	0.3190
0.2877	0.96	300	0.3156

Framework versions

PEFT 0.18.1
Transformers 5.0.0
Pytorch 2.10.0+cu128
Datasets 4.8.4
Tokenizers 0.22.2

Downloads last month: 10

Model tree for sprok-daniel-mozaik/LightOnOCR-2-1B-Math-Handwritten

Base model

lightonai/LightOnOCR-2-1B

Adapter

(7)

this model