Path
Open models on a small budget
Pick a small open model and run it on a free machine. Measure it honestly. Teach it one job it did not know.
- Prepares you for
- It prepares you for the work of running a model cheaply and saying honestly how good it is. That skill sits behind evaluation work, small products, and language projects for Bharat.
- What you learn
- You choose a small open model for one job and run it on a free machine. You measure what it does before changing anything. Then you decide whether teaching it more is worth the money. You build the examples it would learn from. You write down plainly what it still cannot do.
- Why this order
- Measuring comes before changing, always. Otherwise you cannot tell whether the change helped or you just got lucky.
- What an open model really gives you18 minutes
- Pick a small model for one job20 minutes
- Run a model on a free machine22 minutes
- Measure it before you change it22 minutes
- Decide when tuning is worth it22 minutes
- Build the examples it learns from24 minutes
- Tune a model without a big machine24 minutes
- Say plainly what it cannot do20 minutes
Curriculum reviewed: