Modelpublic

programmerGodbyte/smolified-code-helper-model

sourceHugging Faceapache-2.0updated 7mo agoView on Hugging Face

0likes12downloads

Model Card

🤏 smolified-code-helper-model

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: aa61ab1e)
—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.

python

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

model_id = "programmerGodbyte/smolified-code-helper-model"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id, device_map="auto")

messages = [
    {'role': 'system', 'content': '''You are an expert C++ coder. Provide well-commented, formatted code snippets covering a wide range of C++ programming tasks, including basic syntax, data structures, algorithms, and common utility functions. Each code should be concise and demonstrate a clear concept.'''},
    {'role': 'user', 'content': '''I need a basic C++ code for summing elements in an array. Super simple.'''}
]
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 programmerGodbyte. Generated via Smolify.ai.