WikiofAuto
REPRODUCIBLE COMPUTE ARTIFACTS · COLAB T4 & KAGGLE 2×T4 READY

Ready Kaggle & Colab Notebooks + Trained Data Hub

Every notebook is verified on free cloud GPUs (Colab 15GB T4 & Kaggle 30GB Dual-T4), pre-equipped with idle Keep-Alive heartbeats, multi-connection aria2c model pullers, and one-click OpenAI-compatible tunnel endpoints.

KAGGLEKaggle Free 2× NVIDIA T4 (30GB Combined VRAM)
🛡️ Keep-Alive Built-In

Kaggle Dual-T4 (30GB VRAM) Distributed vLLM & Llama-Server 70B Playbook

Unlock Kaggle's free 30GB combined VRAM (2×T4 GPUs, 30 hrs/week) using tensor parallelism to run 32B–70B models at full speed for competitions, bulk synthetic labeling, and remote API serving.

Base Model / Stack:Qwen/Qwen2.5-72B-Instruct-AWQ / Llama-3.3-70B-Instruct-Q4_K_M
Pipeline Method:Tensor Parallelism (--tensor-parallel-size 2) + Ray / llama.cpp Split-Mode
KAGGLEKaggle P100 or Dual T4 GPU
🛡️ Keep-Alive Built-In

Grandmaster Auto-EDA, Agentic Feature Engineering & Optuna Hill-Climbing Ensemble

Competition-winning tabular & multimodal Kaggle playbook: automated adversarial validation, target encoding without leakage, GPU-accelerated GBDT + neural OOF generation, and hill-climbing ensemble blending.

Base Model / Stack:LightGBM + CatBoost + XGBoost + RealMLP + TabNet Stack
Pipeline Method:5-Fold Stratified OOF + Optuna Bayesian Sweep + Nelder-Mead / Hill-Climbing Blender
COLABFree T4 (15GB VRAM) or L4 (24GB VRAM)
🛡️ Keep-Alive Built-In

Qwen 3.8 27B Vision (Abliterated GGUF) — Free Colab OpenAI Server & Cloudflare Tunnel

Run Qwen 3.8 27B Vision (Q4_K_M + mmproj-Q8_0) on a free Google Colab T4 GPU with 16-connection aria2c download, automatic CUDA layer splitting, dual-layer Keep-Alive, and an instant OpenAI-compatible /v1 HTTPS endpoint.

Base Model / Stack:0bserverx/Qwen3.8-27B-Heretic-Abliterated-Uncensored-GGUF
Pipeline Method:llama.cpp CUDA + Dynamic Layer Offload (-ngl) + Cloudflare Reverse Tunnel
COLAB + KAGGLECPU High-RAM or Free T4 GPU
🛡️ Keep-Alive Built-In

Trained Data Pipeline: Magpie Synthetic Reasoning Distillation & MinHash Corpus Scrubber

Complete training-data curation notebook: builds high-signal instruction & reasoning trace datasets (ShareGPT/ChatML), removes near-duplicates via MinHash LSH, scrubs PII/secrets, and formats DPO/GRPO preference pairs.

Base Model / Stack:Any Frontier Teacher -> Domain SLM Student
Pipeline Method:Magpie Zero-Seed Synthesis + MinHash LSH Dedup + LLM-as-Judge Quality Filter + DPO Pair Builder
COLAB + KAGGLEFree T4 (15GB VRAM)
🛡️ Keep-Alive Built-In

Unsloth 2x Faster QLoRA Fine-Tuning → Automated GGUF Export & Ollama Modelfile

End-to-end fine-tuning notebook using Unsloth custom Triton kernels (2.2x faster, 68% less VRAM). Trains on ShareGPT/Alpaca JSONL, merges LoRA adapters, exports Q4_K_M GGUF, and generates an Ollama Modelfile.

Base Model / Stack:Qwen2.5-14B / Llama-3.1-8B / Mistral-Nemo-12B
Pipeline Method:Unsloth 4-bit QLoRA (Rank 16, Triton Kernels) + Direct GGUF Q4_K_M Quantizer
⚡ TRENDING PRODUCTION KITS FOR COLAB & KAGGLE

Turn Free Cloud Notebooks into Production Infrastructure

Browse All Starter Kits →