Spaces:
Running
on
Zero
Running
on
Zero
Commit
Β·
f9ba57a
1
Parent(s):
161eaf7
html title
Browse files
app.py
CHANGED
@@ -190,11 +190,16 @@ def generate_instruction_response():
|
|
190 |
|
191 |
|
192 |
title = """
|
193 |
-
|
194 |
"""
|
195 |
|
196 |
description = """
|
197 |
-
This demo showcases **Magpie**, an innovative approach to generating high-quality data by prompting aligned LLMs with their pre-query templates. Unlike many existing synthetic data generation methods, Magpie doesn't rely on prompt engineering or seed questions for generating synthetic data. Instead, it uses the prompt template of an aligned LLM to generate both
|
|
|
|
|
|
|
|
|
|
|
198 |
|
199 |
As well as providing a demo for the Magpie generations, this Space also allows you to submit a preference rating for the generated data, contributing to a crowdsourced preference dataset!
|
200 |
|
@@ -204,7 +209,6 @@ As well as providing a demo for the Magpie generations, this Space also allows y
|
|
204 |
2. **π¬ Response Generation:** The model generates a response to this instruction.
|
205 |
3. **ππ User Feedback (optional):** Rate the quality of the generated content and contribute to a crowdsourced preference dataset for synthetic dataset.
|
206 |
|
207 |
-
|
208 |
π Find the crowd-generated dataset at [davanstrien/magpie-preference](https://huggingface.co/datasets/davanstrien/magpie-preference). It's updated every minute!
|
209 |
|
210 |
π Learn more about Magpie in the [paper](https://huggingface.co/papers/2406.08464).
|
@@ -214,7 +218,7 @@ As well as providing a demo for the Magpie generations, this Space also allows y
|
|
214 |
|
215 |
# Create the Gradio interface
|
216 |
with gr.Blocks() as iface:
|
217 |
-
gr.
|
218 |
gr.Markdown(description)
|
219 |
|
220 |
# Add a state variable to store the session ID
|
|
|
190 |
|
191 |
|
192 |
title = """
|
193 |
+
<h1>🐦 Magpie Preference</h1>
|
194 |
"""
|
195 |
|
196 |
description = """
|
197 |
+
This demo showcases **[Magpie](https://magpie-align.github.io/)**, an innovative approach to generating high-quality data by prompting aligned LLMs with their pre-query templates. Unlike many existing synthetic data generation methods, Magpie doesn't rely on prompt engineering or seed questions for generating synthetic data. Instead, it uses the prompt template of an aligned LLM to generate both the user query and an LLM response.
|
198 |
+
|
199 |
+
<img src="https://magpie-align.github.io/images/pipeline.png" alt="Magpie Pipeline" width="50%" align="center" />
|
200 |
+
|
201 |
+
*Image Source: [Magpie project page](https://magpie-align.github.io/)*
|
202 |
+
|
203 |
|
204 |
As well as providing a demo for the Magpie generations, this Space also allows you to submit a preference rating for the generated data, contributing to a crowdsourced preference dataset!
|
205 |
|
|
|
209 |
2. **π¬ Response Generation:** The model generates a response to this instruction.
|
210 |
3. **ππ User Feedback (optional):** Rate the quality of the generated content and contribute to a crowdsourced preference dataset for synthetic dataset.
|
211 |
|
|
|
212 |
π Find the crowd-generated dataset at [davanstrien/magpie-preference](https://huggingface.co/datasets/davanstrien/magpie-preference). It's updated every minute!
|
213 |
|
214 |
π Learn more about Magpie in the [paper](https://huggingface.co/papers/2406.08464).
|
|
|
218 |
|
219 |
# Create the Gradio interface
|
220 |
with gr.Blocks() as iface:
|
221 |
+
gr.HTML(title)
|
222 |
gr.Markdown(description)
|
223 |
|
224 |
# Add a state variable to store the session ID
|