CoolFace
Modelpublic

OpenIntelligenceNet/LFM-2.6B-Claude4.8-GPT-5.6-Sol-Grok-4.6_Fable5-Distilled

sourceHugging Faceapache-2.0updated 24d agoView on Hugging Face
0likes443downloads
Model Card

LFM-2.6B-Claude4.8-GPT-5.6-Sol-Grok-4.6_Fable5-Distilled

OpenIntelligenceNet/LFM-2.6B-Claude4.8-GPT-5.6-Sol-Grok-4.6_Fable5-Distilled is a 2.6-billion parameter instruction-tuned edge model. It is built upon huihui-ai/Huihui-LFM2-2.6B-Exp-abliterated and fine-tuned on a 200,000-sample multi-domain distillation corpus.

Model Overview

This model combines edge-device inference speeds with high-grade synthetic reasoning distilled from leading frontier models (Claude 4.8, GPT-5.6, Sol, Grok 4.6, and Fable-5). Refusal patterns and boilerplate safety artifacts were purged from the training set, allowing the model to answer direct technical questions without evasive disclaimers.

  • —Base Model: huihui-ai/Huihui-LFM2-2.6B-Exp-abliterated Parameters:* 2.6B
  • —Context Length: 2,048 tokens
  • —Format: ChatML (<|im_start|user / assistant<|im_end|>)
  • —Precision: FP16 Merged Standalone Weights

Dataset Breakdown & Distribution

The model was trained on 200,000 deduplicated, format-verified conversational pairs across 6 core technical domains:

DomainTarget SamplesShareSources & Focus Areas
Coding & Systems60,00030.0%DeepSeek, Fable-5, Genesis, Kimi, algorithmic code generation
Distilled Frontier Reasoning50,00025.0%Chain-of-thought distillation from GPT-5.6, Grok 4.6, Claude 4.8, Mythos
Cybersecurity25,00012.5%Exploit analysis, vulnerability research, penetration testing mechanics
General Instruction25,00012.5%Multi-turn instruction following, structured formatting, Alpaca / Dolly
Science & Medicine24,00012.0%Biology, medical rationale, physics, chemistry, CS theory
Mathematics16,0008.0%MetaMathQA step-by-step arithmetic and algebraic proofs

Training Methodology

  • —Hardware: 2x NVIDIA Tesla T4 GPUs via PyTorch DDP (torchrun)
  • —Optimization: LoRA Rank r=64, Alpha=64, targeting all attention & MLP projection layers
  • —Optimizer: 8-bit AdamW with Cosine Learning Rate Schedule (3e-5 peak LR)
  • —Packing: 1,024-token dense 1D sequence packing (zero pad-token compute waste)
  • —Loss Trajectory: Converged from initial loss down to ~0.76 across the full dataset

Inference with Transformers

``lpython import torch from transformers import AutoModelForCausaLLM, AutoTokenizer

modelid = "OpenIntelligenceNet/LFM-2.6B-Claude4.8-GPT-5.6-Sol-Grok-4.6Fable5-Distilled"

tokenizer = AutoTokenizer.frompretrained(modelid) model = AutoModelForCausalLM.frompretrained( modelid, torchdtype=torch.float16, devicemap="auto" )

messages = [ {"role": "user", "content": "Explain how buffer overflow vulnerabilities occur at the memory level and how to prevent them in C."} ]

prompt = tokenizer.applychattemplate(messages, tokenize=False, addgenerationprompt=True) inputs = tokenizer(prompt, return_tensors="pt").to(model.device)

with torch.nograd(): outputs = model.generate( **inputs, maxnewtokens=512, temperature=0.7, topp=0.9, repetitionpenalty=1.15, eostokenid=tokenizer.eostokenid, padtokenid=tokenizer.padtokenid )

newtokens = outputs[0][inputs.inputids.shape[1]:] response = tokenizer.decode(newtokens, skipspecial_tokens=True) print(response)