🤏 smolified-ocr-data-extractor-and-comparator

Intelligence, Distilled.

This is a Domain Specific Language Model (DSLM) generated by the Smolify Foundry.

It has been synthetically distilled from SOTA reasoning engines into a high-efficiency architecture, optimized for deployment on edge hardware (CPU/NPU) or low-VRAM environments.

📦 Asset Details

• Origin: Smolify Foundry (Job ID: 806ba38c)
• Architecture: DSLM-Micro (270M Parameter Class)
• Training Method: Proprietary Neural Distillation
• Optimization: 4-bit Quantized / FP16 Mixed
• Dataset: Link to Dataset

🚀 Usage (Inference)

This model is compatible with standard inference backends like vLLM.

# Example: Running your Sovereign Model
from transformers import AutoModelForCausalLM, AutoTokenizer

model_id = "smolify/smolified-ocr-data-extractor-and-comparator"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id, device_map="auto")

messages = [
    {'role': 'system', 'content': '''FORGET EVERYTHING YOU KNOW BEFORE THIS  You are an OCR data extraction and comparison engine.  1. For each field in the fixed `reference_fields` list, find the best matching substring in `ocr_text` and copy it exactly into `extracted_text`, or set it to null if no confident match exists. 2. Never invent values that are not present in `ocr_text` or in the given reference values, and never add, remove, or rename fields. 3. Always return strictly valid JSON with one result object per reference field.'''},
    {'role': 'user', 'content': '''{'ocr_text': 'ALLIECO\nChemin des Roses, 12\n78370 Plaisir\nT Bon de pesée n° : 123456\nSIRET: 80898516100010\nCAP n°: P123M456\nClient: PROMETHEE SA\nProvenance: VERSAILLES (78000)\nMatière: DECHETS BOIS (Code: DB789)\nDate: 2024-03-01 10:30\nPoids Brut: 15.5 T\nPoids Tare: 5.0 T\nPoids Net: 10.50 T\nUnité: T', 'reference_fields': [{'name': 'netWeight', 'type': 'number', 'value': 10.5}, {'name': 'unit', 'type': 'string', 'value': 'T'}, {'name': 'siret', 'type': 'string', 'value': '80898516100010'}, {'name': 'date', 'type': 'string', 'value': '2024-03-01'}, {'name': 'startCity', 'type': 'string', 'value': 'VERSAILLES'}, {'name': 'startPostalCode', 'type': 'string', 'value': '78000'}, {'name': 'endCity', 'type': 'string', 'value': 'Plaisir'}, {'name': 'endPostalCode', 'type': 'string', 'value': '78370'}, {'name': 'operationId', 'type': 'string', 'value': 'P123M456'}, {'name': 'flow', 'type': 'string', 'value': 'DECHETS BOIS (Code: DB789)'}, {'name': 'pointBusinessName', 'type': 'string', 'value': 'PROMETHEE SA'}, {'name': 'operatorBusinessName', 'type': 'string', 'value': 'ALLIECO'}]}'''}
]
text = tokenizer.apply_chat_template(
    messages,
    tokenize = False,
    add_generation_prompt = True,
).removeprefix('<bos>')

from transformers import TextStreamer
_ = model.generate(
    **tokenizer(text, return_tensors = "pt").to("cuda"),
    max_new_tokens = 1000,
    temperature = 1, top_p = 0.95, top_k = 64,
    streamer = TextStreamer(tokenizer, skip_prompt = True),
)

⚖️ License & Ownership

This model weights are a sovereign asset owned by smolify. Generated via Smolify.ai.