Sample Python Code for Model Checkpoints Sample Python Code for Model Checkpoints | Zhiwei Li

Sample Python Code for Model Checkpoints | Generated by AI

Home 2026.04

You have two checkpoints available. Here are sample commands:

From your FineWeb checkpoint (GPT-2 style, BPE tokenizer):

python sample.py \
    --out_dir=out-fineweb \
    --start="The future of artificial intelligence is" \
    --num_samples=3 \
    --max_new_tokens=200

From your Shakespeare char-level checkpoint:

python sample.py \
    --out_dir=out-shakespeare-char \
    --start="To be or not to be" \
    --num_samples=3 \
    --max_new_tokens=200

From a pretrained GPT-2 (no local checkpoint needed):

python sample.py \
    --init_from=gpt2 \
    --start="Once upon a time" \
    --num_samples=2 \
    --max_new_tokens=100

Key flags for sample.py:

Flag Default Description
--num_samples 10 Number of samples to generate
--max_new_tokens 500 Tokens per sample
--temperature 0.8 Higher = more random
--top_k 200 Top-k sampling cutoff
--start "\n" Prompt string (or FILE:prompt.txt)
--device cuda Use cpu if no GPU
Back Donate