File size: 774 Bytes
59d2e5c
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2c14e6e
 
 
 
 
 
 
 
 
afcfe6e
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
---
dataset_info:
  features:
  - name: questions
    dtype: string
  - name: reasoning
    dtype: string
  - name: answers
    dtype: string
  splits:
  - name: test
    num_bytes: 79373605
    num_examples: 388856
  download_size: 30411606
  dataset_size: 79373605
configs:
- config_name: default
  data_files:
  - split: test
    path: data/test-*
---

# MNLP Quantization Dataset

This dataset is a copy of the MCQA dataset used during our project and reused here to calibrate our quantized model using post-training quantization (PTQ).

## Origin
Originally hosted at: [`hssawhney/MNLP-Dataset`](https://huggingface.co/datasets/hssawhney/MNLP-Dataset)

## Use
Used to calibrate our quantized W8A8 MCQA model. The quantization setup used 512 examples from the test set.