Spaces:
Sleeping
Sleeping
Commit
·
608d7cd
1
Parent(s):
1a3ed87
Update space
Browse files
app.py
CHANGED
|
@@ -1,62 +1,106 @@
|
|
| 1 |
import gradio as gr
|
| 2 |
from huggingface_hub import InferenceClient
|
|
|
|
|
|
|
| 3 |
|
| 4 |
-
""
|
| 5 |
-
For more information on `huggingface_hub` Inference API support, please check the docs: https://huggingface.co/docs/huggingface_hub/v0.22.2/en/guides/inference
|
| 6 |
-
"""
|
| 7 |
-
client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")
|
| 8 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 9 |
|
| 10 |
-
|
| 11 |
-
message,
|
| 12 |
-
history: list[tuple[str, str]],
|
| 13 |
-
system_message,
|
| 14 |
-
max_tokens,
|
| 15 |
-
temperature,
|
| 16 |
-
top_p,
|
| 17 |
-
):
|
| 18 |
-
messages = [{"role": "system", "content": system_message}]
|
| 19 |
|
| 20 |
-
|
| 21 |
-
|
| 22 |
-
|
| 23 |
-
|
| 24 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 25 |
|
| 26 |
-
|
| 27 |
|
| 28 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 29 |
|
| 30 |
-
|
| 31 |
-
messages,
|
| 32 |
-
max_tokens=max_tokens,
|
| 33 |
-
stream=True,
|
| 34 |
-
temperature=temperature,
|
| 35 |
-
top_p=top_p,
|
| 36 |
-
):
|
| 37 |
-
token = message.choices[0].delta.content
|
| 38 |
|
| 39 |
-
|
| 40 |
-
|
|
|
|
|
|
|
|
|
|
| 41 |
|
|
|
|
|
|
|
| 42 |
|
| 43 |
-
|
| 44 |
-
|
| 45 |
-
""
|
| 46 |
-
|
| 47 |
-
|
| 48 |
-
|
| 49 |
-
|
| 50 |
-
|
| 51 |
-
|
| 52 |
-
gr.Slider(
|
| 53 |
-
minimum=0.1,
|
| 54 |
-
maximum=1.0,
|
| 55 |
-
value=0.95,
|
| 56 |
-
step=0.05,
|
| 57 |
-
label="Top-p (nucleus sampling)",
|
| 58 |
-
),
|
| 59 |
-
],
|
| 60 |
)
|
| 61 |
|
| 62 |
|
|
|
|
| 1 |
import gradio as gr
|
| 2 |
from huggingface_hub import InferenceClient
|
| 3 |
+
import ast
|
| 4 |
+
from graphviz import Digraph
|
| 5 |
|
| 6 |
+
client = InferenceClient("Qwen/Qwen2.5-72B-Instruct")
|
|
|
|
|
|
|
|
|
|
| 7 |
|
| 8 |
+
def sampling(num_samples, num_associations):
|
| 9 |
+
outputs = ast.literal_eval(client.chat.completions.create(
|
| 10 |
+
messages=[
|
| 11 |
+
{"role": "system", "content": "generate one json object, no explanation or additional text, use the following structure:\n"
|
| 12 |
+
"words: []\n"
|
| 13 |
+
f"{num_samples} samples in a list"
|
| 14 |
+
},
|
| 15 |
+
{"role": "user",
|
| 16 |
+
"content": f"synthesize {num_samples} random but widespread words for semantic modeling"},
|
| 17 |
+
],
|
| 18 |
+
response_format={
|
| 19 |
+
"type": "json",
|
| 20 |
+
"value": {
|
| 21 |
+
"properties": {
|
| 22 |
+
"words": {"type": "array", "items": {"type": "string"}},
|
| 23 |
+
}
|
| 24 |
+
}
|
| 25 |
+
},
|
| 26 |
+
stream=False,
|
| 27 |
+
max_tokens=1024,
|
| 28 |
+
temperature=0.7,
|
| 29 |
+
top_p=0.1
|
| 30 |
+
).choices[0].get('message')['content'])
|
| 31 |
|
| 32 |
+
fields = {}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 33 |
|
| 34 |
+
for word in outputs['words']:
|
| 35 |
+
fields[word] = ast.literal_eval(client.chat.completions.create(
|
| 36 |
+
messages=[
|
| 37 |
+
{"role": "system", "content": 'generate one json object, no explanation or additional text, use the following structure:\n'
|
| 38 |
+
'associations: []'
|
| 39 |
+
},
|
| 40 |
+
{"role": "user",
|
| 41 |
+
"content": f"synthesize {num_associations} associations for the word {word}"},
|
| 42 |
+
],
|
| 43 |
+
response_format={
|
| 44 |
+
"type": "json",
|
| 45 |
+
"value": {
|
| 46 |
+
"properties": {
|
| 47 |
+
"associations": {"type": "array", "items": {"type": "string"}}
|
| 48 |
+
}
|
| 49 |
+
}
|
| 50 |
+
},
|
| 51 |
+
stream=False,
|
| 52 |
+
max_tokens=2000,
|
| 53 |
+
temperature=0.7,
|
| 54 |
+
top_p=0.1
|
| 55 |
+
).choices[0].get('message')['content'])
|
| 56 |
|
| 57 |
+
triplets = []
|
| 58 |
|
| 59 |
+
for cluster in fields:
|
| 60 |
+
for association in fields[cluster]['associations']:
|
| 61 |
+
triplets.append(ast.literal_eval(client.chat.completions.create(
|
| 62 |
+
messages=[
|
| 63 |
+
{"role": "system", "content": "generate one json object, no explanation or additional text, use the following structure:\n"
|
| 64 |
+
"properties: [subject, predicate, object]\n"
|
| 65 |
+
"use chain-of-thought for predictions"
|
| 66 |
+
},
|
| 67 |
+
{"role": "user",
|
| 68 |
+
"content": f"form triplet based on semantics: generate predicate between the word {cluster} (subject) and the word {association} (object); return list with [subject, predicate, object]"},
|
| 69 |
+
],
|
| 70 |
+
response_format={
|
| 71 |
+
"type": "json",
|
| 72 |
+
"value": {
|
| 73 |
+
"properties": {
|
| 74 |
+
"properties": {"type": "array", "items": {"type": "string"}}
|
| 75 |
+
}
|
| 76 |
+
}
|
| 77 |
+
},
|
| 78 |
+
stream=False,
|
| 79 |
+
max_tokens=128,
|
| 80 |
+
temperature=0.7,
|
| 81 |
+
top_p=0.1
|
| 82 |
+
).choices[0].get('message')['content']))
|
| 83 |
|
| 84 |
+
dot = Digraph(comment=f'SynthNet, {num_samples} samples, {num_associations} associations', graph_attr={'rankdir': 'LR'})
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 85 |
|
| 86 |
+
for entry in triplets:
|
| 87 |
+
source, label, target = entry['properties']
|
| 88 |
+
dot.node(source, source)
|
| 89 |
+
dot.node(target, target)
|
| 90 |
+
dot.edge(source, target, label=label)
|
| 91 |
|
| 92 |
+
dot.render('synthnet', format='png')
|
| 93 |
+
return 'synthnet.png'
|
| 94 |
|
| 95 |
+
demo = r.Interface(
|
| 96 |
+
inputs=[
|
| 97 |
+
gr.Slider(minimum=1, maximum=256, label="Number of Samples"),
|
| 98 |
+
gr.Slider(minimum=1, maximum=256, label="Number of Associations to each Sample"),
|
| 99 |
+
],
|
| 100 |
+
fn=sampling,
|
| 101 |
+
outputs=gr.Image(type="filepath"),
|
| 102 |
+
title="SynthNet",
|
| 103 |
+
description="Select a number of samples and assiciations to each sample to generate a graph.",
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 104 |
)
|
| 105 |
|
| 106 |
|