About Lab Eval
conversational ai: speech processing and synthesis
A candidate’s competence in practical programming skills on the subject shall be evaluated on the given date. The procedure is simple,
- Create a git repo;
- Init Colab Notebook;
- Solve the given problem; and
- Commit back to repo;
- Submit the Repo URL and Commit ID to the Form.
- The eval is a timed exercise. Mind your time accordingly.
- PS: The form allows edits.
Remember:
- All exercise(s) shall be solved in (Colab) python
notebook(s), committed to Github using
@thapar.eduaccount. - Only a Github Repo link and commit id shall be submit using the Google Form. Any attachments are not allowed.
Lab Eval Registration
Register for Lab Eval using your Name, Email
(@thapar.edu), Roll No., Group and Viva-voce time
preference.
Procedure
Practice using Colab with Github, as follows, so as to relieve yourself of the burden during the event.
Setup a Git Repo
- Login to Github, using
@thapar.eduaccount. - Create a new repo named
<ROLLNO>-SESS_LE1, e.g.102938478-SESS_LE1; so that the linkhttps://github.com/<USER>/<ROLLNO>-SESS_LE1is a valid Github Repo. The repo may be private to avoid - Grant read access to your instructor(s), namely
bv.raghav@thapar.edu.
Init your notebook
- Open this colab.
- Login using
@thapar.eduaccount. - Rename the file as
<ROLLNO>-<NAME>.ipynb, and remove any whitespace within e.g.102938478-BvRaghav.ipynb - Commit it to the root folder of your repository,
using “Save a copy in GitHub.” [Read more…]
- The app will ask for authorisation.
- Remember to check the “Include a link to Colab” option while saving.
- PS: The authorisation may be revoked anytime after the eval.
- The last step should automatically redirect to the
latest version of this notebook, henceforth referred
to as NB, on your GitHub Repo. Click on badge
“Open in Colab.”
- If the badge does not show up, copy URL from the browser’s URL-bar, and load your private notebook.
- Edit your response in NB.
- To commit again, “Save a copy in GitHub” from within NB.
Problem/Solution
On the day of eval, the candidate will be assigned a question (e.g. the sample question further in the document).
- Copy-paste the question verbatim into the
## Questiontext block of the NB. PS: Keep the heading intact. - Write the code.
- Commit using “Save a copy in GitHub” from within NB.
- Submit your response as your Github Link and Git Commit into this Response Form.
Sample Lab Eval Question
Given the dimensionality d of inputs, a sequence H
of channel size for each hidden layer, and number of
classes C, define a function to
- Take as argument a vector of raw inputs
x; - Define a neural network classifier with
dinput channels,len(H)hidden layers, each subsequent layer bearing[h1,h2,...]channels finally resulting inClogits corresponding to each class; - Activate each intermediate layer with \(\tanh\) activation; and
- Return the logits.
The function shall be tested for consistency, correctness and efficiency as applicable.
In all cases the input will be in batches, where batch-size is controlled by the end-user.
Use PyTorch/Tensorflow for implementation. Use of internet resources is dicouraged in the interest of time, though not prohibited.