#!/bin/bash

# This script is the "Best ChatGPT clone that $100 can buy",
# It is designed to run in ~4 hours on 8XH100 node

# Example launch in a screen session (because the run takes ~4 hours):
# screen -L -Logfile speedrun.log -S speedrun bash speedrun.sh

git clone https://github.com/karpathy/nanochat.git
cd nanochat

# Default intermediate artifacts directory is in ~/.cache/nanochat
export OMP_NUM_THREADS=1
export NANOCHAT_BASE_DIR="$HOME/.cache/nanochat"
export WANDB_DISABLED=true

# -----------------------------------------------------------------------------
# Python venv setup with uv

# install uv (if not already installed)
command -v uv &> /dev/null || curl -LsSf https://astral.sh/uv/install.sh | sh
source $HOME/.local/bin/env
# create a .venv local virtual environment (if it doesn't exist)
[ -d ".venv" ] || uv venv
# install the repo dependencies
uv sync
# activate venv so that `python` uses the project's venv instead of system python
source .venv/bin/activate

# -----------------------------------------------------------------------------
# During the course of the run, we will be writing markdown reports to the report/
# directory in the base dir. This command clears it out and writes a header section
# with a bunch of system info and a timestamp that marks the start of the run.
python -m nanochat.report reset

# -----------------------------------------------------------------------------
# Tokenizer

# Install Rust / Cargo
curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh -s -- -y
source "$HOME/.cargo/env"

# Build the rustbpe Tokenizer
uv run maturin develop --release --manifest-path rustbpe/Cargo.toml

# Download the first ~2B characters of pretraining dataset
# look at dev/repackage_data_reference.py for details on how this data was prepared
# each data shard is ~250M chars
# so we download 2e9 / 250e6 = 8 data shards at this point
# each shard is ~100MB of text (compressed), so this is about ~800MB of data on disk
python -m nanochat.dataset -n 8
# Immediately also kick off downloading more shards in the background while tokenizer trains
# See comment below for why 240 is the right number here
python -m nanochat.dataset -n 240 &
DATASET_DOWNLOAD_PID=$!
# train the tokenizer with vocab size 2**16 = 65536 on ~2B characters of data
python -m scripts.tok_train --max_chars=2000000000
# evaluate the tokenizer (report compression ratio etc.)
python -m scripts.tok_eval

# -----------------------------------------------------------------------------
# Base model (pretraining)

# Download the eval_bundle from s3 to evaluate CORE metric during training (~162MB)
EVAL_BUNDLE_URL=https://karpathy-public.s3.us-west-2.amazonaws.com/eval_bundle.zip
if [ ! -d "$NANOCHAT_BASE_DIR/eval_bundle" ]; then
    curl -L -o eval_bundle.zip $EVAL_BUNDLE_URL
    unzip -q eval_bundle.zip
    rm eval_bundle.zip
    mv eval_bundle $NANOCHAT_BASE_DIR
fi

# The d20 model is 561M parameters.
# Chinchilla says #tokens = 20X #params, so we need 561e6 * 20 = 11.2B tokens.
# Assume our tokenizer is 4.8 chars/token, this is 11.2B * 4.8 ~= 54B chars.
# At 250M chars/shard, this is 54B / 250M ~= 216 shards needed for pretraining.
# Round up to 240 for safety. At ~100MB/shard, this downloads ~24GB of data to disk.
# (The total number of shards available in the entire dataset is 1822.)
echo "Waiting for dataset download to complete..."
wait $DATASET_DOWNLOAD_PID

# pretrain the d20 model
torchrun --standalone --nproc_per_node=8 -m scripts.base_train -- --depth=20
# evaluate the model on a larger chunk of train/val data and draw some samples
torchrun --standalone --nproc_per_node=8 -m scripts.base_loss
# evaluate the model on CORE tasks
torchrun --standalone --nproc_per_node=8 -m scripts.base_eval

# -----------------------------------------------------------------------------
# Midtraining (teach the model conversation special tokens, tool use, multiple choice)

# run midtraining and eval the model
torchrun --standalone --nproc_per_node=8 -m scripts.mid_train
torchrun --standalone --nproc_per_node=8 -m scripts.chat_eval -- -i mid

# -----------------------------------------------------------------------------
# Supervised Finetuning (domain adaptation to each sequence all by itself per row)

# train sft and re-eval right away (should see a small bump)
torchrun --standalone --nproc_per_node=8 -m scripts.chat_sft
torchrun --standalone --nproc_per_node=8 -m scripts.chat_eval -- -i sft

# -----------------------------------------------------------------------------
# Generate the full report by putting together all the sections
# report.md is the output and will be copied to current directory for convenience
python -m nanochat.report generate