Playground
Pick a model and a training step, then give it a prompt. Slide back to step 0 to see where it started. GPT-2 124M on 10B tokens, 8× H100 (same recipe, rented for 90 min): curves and timeline
GPT-2 124M on 10B tokens, 8× H100 (same recipe, rented for 90 min)
These are base models: they continue text, they don't answer questions. Chat comes with Phase 2's fine-tuning.