Propose a question for fast distributed framework (#27)

- Propose a question for fast distributed framework (ea0b67933658cdcc16807dd6ac0877909c2b6c02)
- Update app.py (1c9f9bafaf54880d88b9c4996316a43893edbd56)
- Use real values so that the answer is an actual value (69e6c574c14aee05ee51d8883699e75dedca8f60)
- Update app.py (12ac04b43efee8de85d619f5df24142e33603241)


Co-authored-by: Thomas Wang <TimeRobber@users.noreply.huggingface.co>

app.py

@@ -115,6 +115,16 @@ For some resources, you may want to check out:
 * <a href="https://huggingface.co/docs/simulate/main/en/tutorials/running_the_simulation#running-the-simulation" simulation stepping <a> for running the simulation.
 """
 
+fast_distributed_framework_question = """
+We are going to understand how many operations does a matrix multiplication hold.
+<br/>
+<br/>
+Let N=128, M=256, K=64
+Let A,B two matrices of size 256x64 and 128x64 respectively. When computing the matrix multiplication of A and (B^T), how many scalar multiplications are done? How many scalar additions are done? Please answer in the following format:
+multiplications: {YOUR_ANSWER_AS_A_SINGLE_NUMBER}
+additions: {YOUR_ANSWER_AS_A_SINGLE_NUMBER}
+"""
+
 internships = {
     'Accelerate': default_question,
     'Skops & Scikit-Learn': skops_question,
@@ -134,7 +144,7 @@ internships = {
     "Benchmarking transformers on various AI hardware accelerators": default_question,
     "AI Art Tooling Residency": default_question,
     "Datasets for Large Language Models": default_question,
-    "Fast Distributed Training Framework": default_question,
+    "Fast Distributed Training Framework": fast_distributed_framework_question,
 }