Original Academy
Legacy mirror · noindex · upstream phase 10
LLMs from Scratch
Build, train, and understand large language models.
Provenance: this phase outline and its lesson readings are preserved from ai-engineering-from-scratch by Rohit Ghumare under the MIT License. InterviewsVector does not claim authorship. These archive pages remain available for old links and progress, but are excluded from indexing.
Attributed readings
- 01Tokenizers: BPE, WordPiece, SentencePieceYour LLM does not read English. It reads integers. The tokenizer decides whether those integers carry meaning or waste it.
- 02Building a Tokenizer from ScratchLesson 01 gave you a toy. This lesson gives you a weapon.
- 03Data Pipelines for Pre-TrainingThe model is a mirror. It reflects whatever data you feed it. Feed it garbage, it reflects garbage with perfect fluency.
- 04Pre-Training a Mini GPT (124M)GPT-2 Small has 124 million parameters. That's 12 transformer layers, 12 attention heads, and 768-dimensional embeddings. You can train it from scratch on a single GPU in a few …
- 05Distributed Training, FSDP, DeepSpeedYour 124M model trained on one GPU. Now try 7 billion parameters. The model doesn't fit in memory. The data takes weeks on a single machine. Distributed training isn't optional …
- 06Instruction Tuning — SFTA base model predicts the next token. That's it. It doesn't follow instructions, answer questions, or refuse harmful requests. SFT is the bridge between a token predictor and a …
- 07RLHF — Reward Model + PPOSFT teaches the model to follow instructions. But it doesn't teach the model which response is BETTER. Two grammatically correct, factually accurate answers can differ enormousl…
- 08DPO — Direct Preference OptimizationRLHF works. It also requires training three models (SFT, reward model, policy), managing PPO's instability, and tuning a KL penalty. DPO asks: what if you could skip all of that…
- 09Constitutional AI & Self-ImprovementRLHF needs humans in the loop. Constitutional AI replaces most of them with the model itself. Write a list of principles, have the model critique its own outputs against those p…
- 10Evaluation — Benchmarks, EvalsGoodhart's Law: when a measure becomes a target, it ceases to be a good measure. Every frontier lab games benchmarks. MMLU scores go up while models still can't reliably count t…
- 11Quantization: INT8, GPTQ, AWQ, GGUFA 70B model in FP16 needs 140GB. Two A100s just for weights. Quantize to FP8: one 80GB GPU. INT4: a MacBook.
- 12Inference OptimizationTwo phases define LLM inference. Prefill processes your prompt in parallel -- compute-bound. Decode generates tokens one at a time -- memory-bound. Every optimization targets on…
- 13Building a Complete LLM PipelineEverything from Lessons 01 to 12 is one stage of one pipeline. This lesson is the scaffold that turns those stages into a single end-to-end run: tokenize, pre-train, scale, SFT,…
- 14Open Models: Architecture WalkthroughsYou built a GPT-2 Small from scratch in Lesson 04. Frontier open models in 2026 are the same family with five or six concrete changes. RMSNorm instead of LayerNorm. SwiGLU inste…
- 15Speculative Decoding and EAGLE-3Phase 7 · Lesson 16 proved the math: the Leviathan rejection rule preserves the verifier's distribution exactly. This lesson is the training-stack view of 2026 production specul…
- 16Differential Attention (V2)Softmax attention spreads a small amount of probability over every non-matching token. Over 100k tokens that noise adds up and drowns the signal. Differential Transformer (Ye et…
- 17Native Sparse Attention (DeepSeek NSA)At 64k tokens, attention eats 70-80% of decode latency. Every open-model lab has a plan to fix it. DeepSeek's NSA (ACL 2025 best paper) is the one that stuck: three parallel att…
- 18Multi-Token Prediction (MTP)Every autoregressive LLM from GPT-2 to Llama 3 trains on one loss per position: predict the next token. DeepSeek-V3 added a second loss per position: predict the token after tha…
- 19DualPipe ParallelismDeepSeek-V3 was trained on 2,048 H800 GPUs with MoE experts scattered across nodes. Cross-node expert all-to-all communication cost 1 GPU-hour of comm for every 1 GPU-hour of co…
- 20DeepSeek-V3 Architecture WalkthroughPhase 10 · Lesson 14 named the six architectural knobs every open model turns. DeepSeek-V3 (December 2024, 671B parameters total, 37B active) turns all six and adds four more: M…
- 21Jamba — Hybrid SSM-TransformerState space models (SSMs) and transformers want different things. Transformers buy quality via attention at quadratic cost. SSMs buy linear-time inference and constant memory vi…
- 22Async and Hogwild! InferenceSpeculative decoding (Phase 10 · 15) parallelizes tokens within one sequence. Multi-agent frameworks parallelize across whole sequences but force explicit coordination (voting, …
- 23Speculative Decoding and EAGLEA frontier LLM generating one token requires a full forward pass over billions of parameters. That forward pass is massively over-provisioned: most of the time a much smaller mo…
- 24Gradient Checkpointing and Activation RecomputationBackprop keeps every intermediate activation. At 70B parameters and 128K context that is 3 TB of activations per rank. Checkpointing trades FLOPs for memory: recompute instead o…