Spaces:
Paused
Paused
Update app.py
Browse files
app.py
CHANGED
@@ -3,7 +3,12 @@ import torch.nn as nn
|
|
3 |
from torch.nn import functional as F
|
4 |
import tiktoken
|
5 |
import gradio as gr
|
6 |
-
|
|
|
|
|
|
|
|
|
|
|
7 |
# Define the model architecture
|
8 |
class GPTConfig:
|
9 |
def __init__(self):
|
@@ -128,8 +133,8 @@ import gradio as gr
|
|
128 |
|
129 |
# [Your existing model code remains unchanged]
|
130 |
|
131 |
-
#
|
132 |
-
def generate_text(prompt, max_length=100, temperature=0.7, top_k=50):
|
133 |
input_ids = torch.tensor(enc.encode(prompt)).unsqueeze(0)
|
134 |
generated = []
|
135 |
|
@@ -150,13 +155,14 @@ def generate_text(prompt, max_length=100, temperature=0.7, top_k=50):
|
|
150 |
|
151 |
if next_token.item() == enc.encode('\n')[0] and len(generated) > 20:
|
152 |
break
|
|
|
|
|
153 |
|
154 |
-
#
|
155 |
-
def gradio_generate(prompt, max_length, temperature, top_k):
|
156 |
output = ""
|
157 |
-
for token in generate_text(prompt, max_length, temperature, top_k):
|
158 |
output += token
|
159 |
-
time.sleep(0.05) # Simulate typing effect
|
160 |
yield output
|
161 |
|
162 |
# Custom CSS for the animation effect
|
|
|
3 |
from torch.nn import functional as F
|
4 |
import tiktoken
|
5 |
import gradio as gr
|
6 |
+
import torch
|
7 |
+
import torch.nn as nn
|
8 |
+
from torch.nn import functional as F
|
9 |
+
import tiktoken
|
10 |
+
import gradio as gr
|
11 |
+
import asyncio
|
12 |
# Define the model architecture
|
13 |
class GPTConfig:
|
14 |
def __init__(self):
|
|
|
133 |
|
134 |
# [Your existing model code remains unchanged]
|
135 |
|
136 |
+
# Modify the generate_text function to be asynchronous
|
137 |
+
async def generate_text(prompt, max_length=100, temperature=0.7, top_k=50):
|
138 |
input_ids = torch.tensor(enc.encode(prompt)).unsqueeze(0)
|
139 |
generated = []
|
140 |
|
|
|
155 |
|
156 |
if next_token.item() == enc.encode('\n')[0] and len(generated) > 20:
|
157 |
break
|
158 |
+
|
159 |
+
await asyncio.sleep(0.05) # Use asyncio.sleep instead of time.sleep
|
160 |
|
161 |
+
# Modify the gradio_generate function to be asynchronous
|
162 |
+
async def gradio_generate(prompt, max_length, temperature, top_k):
|
163 |
output = ""
|
164 |
+
async for token in generate_text(prompt, max_length, temperature, top_k):
|
165 |
output += token
|
|
|
166 |
yield output
|
167 |
|
168 |
# Custom CSS for the animation effect
|