Write Clean PyTorch Research Code | Dataford Interview Questions

Write clean, bug-free Python/PyTorch code for a research model.

Implement research_model(sequences, mask, weight, bias). Each sequence has shape [length, features], and mask marks valid tokens with 1 or padding with 0. Compute the masked mean feature vector for every batch item, then apply the affine transformation weight @ pooled + bias, where weight has shape [outputs, features]. Return a nested Python list of floats.

Constraints

1 <= len(sequences) <= 1000
1 <= sequence length <= 1000
1 <= feature count <= 100
1 <= output count <= 100
All sequences have the same feature count
mask values are 0 or 1
weight has shape [output count, feature count]
bias has length output count

Function Signature

def research_model(sequences, mask, weight, bias):

Your question is Write Clean PyTorch Research Code. Start with the requirements in the Question tab.

Run and submit as often as you like. When you're ready, talk me through your approach or go straight to the code.

Problem

Write clean, bug-free Python/PyTorch code for a research model.

Constraints

1 <= len(sequences) <= 1000
1 <= sequence length <= 1000
1 <= feature count <= 100
1 <= output count <= 100
All sequences have the same feature count
mask values are 0 or 1
weight has shape [output count, feature count]
bias has length output count

Function Signature

def research_model(sequences, mask, weight, bias):

Your question is Write Clean PyTorch Research Code. Start with the requirements in the Question tab.

Run and submit as often as you like. When you're ready, talk me through your approach or go straight to the code.