Debug a broken GPT training notebook containing NaNs, a mask bug, and shifted labels

Hard

Correctness, Debugging, and Performance Experiments5 / 8

Chapter 265 / 8

Correctness, Debugging, and Performance Experiments

Hard

Target interface

training_step(model, input_ids, attn_mask_bias)

Complete the starter code in the editor. Any correct approach is accepted.

Hints0 / 2

Reference solution

Your own code stays in the editor.

solution.pyPython

Local draft

import torch, torch.nn.functional as F

def training_step(model, input_ids, attn_mask_bias):
    raise NotImplementedError

Local draft

⌘/Ctrl↵to submit

AI review

Ready when you are

Submit your solution for a score and feedback on correctness, edge cases, and clarity. Any correct approach is accepted.

Chapter 265 / 8

Correctness, Debugging, and Performance Experiments

Hard

Target interface

training_step(model, input_ids, attn_mask_bias)

Complete the starter code in the editor. Any correct approach is accepted.

Hints0 / 2

Reference solution

Your own code stays in the editor.