System online—86 notebooks loaded

LLM
DOJO.

A structured curriculum for understanding and training large language models. Foundations to production. No prerequisites beyond Python and PyTorch.

83
Notebooks
7
Stages
Free
Forever
T4 GPU
Tested
Start from zero →View roadmapUse with Claude Code

Curriculum — 8 stages

Full roadmap
Stage 007 notebooks

Foundations & Environment

Set up your environment, understand transformers from scratch, and master data preparation fundamentals.

open_stage() →
Stage 0110 notebooks

Full Model Fine-Tuning

Fine-tune complete transformer models for classification and generation. Build custom training loops with mixed precision.

open_stage() →
Stage 0210 notebooks

Parameter-Efficient Fine-Tuning

Master LoRA, QLoRA, adapters, and prompt tuning — train LLMs with 100x fewer parameters.

open_stage() →
Stage 0310 notebooks

Advanced Optimization

FlashAttention, DeepSpeed ZeRO, FSDP, gradient checkpointing, and instruction tuning at scale.

open_stage() →
Stage 0410 notebooks

Alignment & Specialized Techniques

RLHF, DPO, Constitutional AI, reward models, and safety evaluation for aligned LLMs.

open_stage() →
Stage 0510 notebooks

Custom Kernels & Production

Write CUDA/Triton kernels, master quantization, implement speculative decoding, and deploy with vLLM.

open_stage() →
Stage 0620 notebooks

LLM Inference Optimization

Profile, optimize, and deploy LLM inference at scale — from KV cache to quantization to multi-GPU serving.

open_stage() →
Stage 079 notebooks

2024–2025 Techniques

GRPO reasoning models, ORPO/KTO alignment, Unsloth acceleration, SGLang serving, synthetic data pipelines, model merging, and standardized evaluation.

open_stage() →

What this is

An open-source curriculum for learning LLM fine-tuning and inference optimization end-to-end. Every notebook runs on Google Colab free tier (T4 GPU, 15 GB VRAM).

Starts with transformer architecture from scratch. Ends with CUDA kernels, GPTQ/AWQ quantization, and production inference with vLLM.

What this isn't

  • —Not hand-holding
  • —Not paid
  • —Not pseudocode — every cell runs
  • —Not theory-only — every concept has a working implementation

Start anywhere. Go as deep as you need.

83 notebooks. 0 cost.

Begin →