LoRA: Trainable Parameter Count
Beginner
The problem, from first principles
LoRA's appeal is the parameter count. A full weight matrix has d_in times d_out entries; the low-rank update has only r times the sum of the two dimensions.
From theory to code
Implement lora_param_count(d_in, d_out, r), the trainable parameters of one LoRA adapter.
Constraints
- Returns an integer.
Hints
Hint 1
Add the two dimensions and multiply by the rank.
Keywords
research-paperssystemsfine-tuningadapters
Python
Run your code to see results
No results yet
Run your code to check it against the examples, or submit to run every test.