Thinking Machines Tinker
A model fine-tuning platform and SDK released by Thinking Machines. It provides low-level primitives for controlling gradients and supports LoRA fine-tuning.
Timeline (last 90 days)
SlothOps uses Thinking Machines' Tinker SDK to perform LoRA fine-tuning on meta-llama/Llama-3-8B-Instruct with rank 16 and alpha 32, applied to the four projection layers q/v/k/o.
One of the core motivations for open-sourcing SlothOps is the data boundary: enterprise chore data (emails, calendar discussions, code diffs, internal memos) is intellectual property, and using open-weight models and the Backboard memory layer can prevent content from being sent into closed-source models for training.
All Facts (2)
SlothOps uses Thinking Machines' Tinker SDK to perform LoRA fine-tuning on meta-llama/Llama-3-8B-Instruct with rank 16 and alpha 32, applied to the four projection layers q/v/k/o.
50%UnverifiedOne of the core motivations for open-sourcing SlothOps is the data boundary: enterprise chore data (emails, calendar discussions, code diffs, internal memos) is intellectual property, and using open-weight models and the Backboard memory layer can prevent content from being sent into closed-source models for training.
50%