Tinker - Thinking Machines Lab

Tinker is a training API for researchers and developers.

Tinker is a training API for researchers Control every aspect of model training and fine-tuning while we handle the infrastructure. Your ideas in four functions forward_backward Perform a forward pass and a backward pass, accumulating the gradient. optim_step Update weights based on the accumulated gradient. sample Generate tokens for interaction, evaluation, or RL actions. save_state Save training progress for resumption. How teams use Tinker Supported models QWEN Dense MoE Dense Dense Dense Dense MoE MoE Dense Explore more Tinker uses LoRA LoRA fine-tunes models by…

saved by

• Katherine Huang
• Linda
• Lydia Nottingham

related reading

• Anatomy of a Modern Finetuning APIbenanderson.work
• LoRA Without Regret - Thinking Machines Labthinkingmachines.ai
• The Thinking Machines Tinker API is good news for AI control and security — LessWronglesswrong.com
• Inkling: Our Open-Weights Model - Thinking Machines Labthinkingmachines.ai
• Tinker: Call for Community Projects - Thinking Machines Labthinkingmachines.ai
• Announcing Safety Research Grantsthinkingmachines.ai
• [2106.09685] LoRA: Low-Rank Adaptation of Large Language Modelsarxiv.org
• Training LLMs to Predict World Events (Guest Post with Mantic) - Thinking Machines Labthinkingmachines.ai
• tinker-nomicstinker-nomics.vercel.app
• On-Policy Distillation - Thinking Machines Labthinkingmachines.ai
• Tutorials - Tinker Documentationtinker-docs.thinkingmachines.ai
• I Let AI Agents Train Their Own Models. Here's What Actually Happened. | Hamza Mostafahamzamostafa.com