Training guide

• › Python 3.9+ with pip
• › PyTorch 2.0+ with CUDA support
• › 16 GB GPU minimum, 40 GB+ for larger models
• › Training data in text format (one document per line recommended)

# Unzip your export
unzip Otter_export.zip
cd Otter_model

# Install dependencies
pip install -r requirements.txt

# Prepare your data
curl -O https://raw.githubusercontent.com/karpathy/char-rnn/master/data/tinyshakespeare/input.txt
mv input.txt data/train.txt

# Start training
python train.py --data_dir data --epochs 10

Free GPU access with limitations. Good for short experiments.

!unzip Otter_export.zip
%cd Otter_model
!pip install -r requirements.txt

!curl -O https://raw.githubusercontent.com/karpathy/char-rnn/master/data/tinyshakespeare/input.txt
!mkdir -p data && mv input.txt data/train.txt

!python train.py --data_dir data --epochs 5 --batch_size 4

→ Colab free tier has 90-min session timeouts and T4 (16 GB) memory. Use Colab Pro for longer runs.

For serious runs, rent an A100 or H100. Lambda Labs, RunPod, and Vast.ai are reliable.

ssh ubuntu@<instance-ip>
scp -r ./data ubuntu@<instance-ip>:~/data
scp Otter_export.zip ubuntu@<instance-ip>:~/
unzip Otter_export.zip && cd Otter_model

pip install -r requirements.txt
python train.py \
  --data_dir ~/data \
  --epochs 100 \
  --batch_size 32 \
  --learning_rate 3e-4 \
  --gradient_accumulation_steps 4

For richer telemetry, integrate Weights & Biases:

pip install wandb
wandb login
python train.py --data_dir data --use_wandb --wandb_project my-llm