Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
@@ -1,5 +1,10 @@
|
|
1 |
import os
|
2 |
-
|
|
|
|
|
|
|
|
|
|
|
3 |
if os.environ.get("SPACES_ZERO_GPU") is not None:
|
4 |
import spaces
|
5 |
else:
|
@@ -14,11 +19,6 @@ else:
|
|
14 |
def fake_gpu():
|
15 |
pass
|
16 |
|
17 |
-
import numpy as np
|
18 |
-
import torch
|
19 |
-
import gradio as gr
|
20 |
-
from transformers import AutoModelForCausalLM, AutoTokenizer
|
21 |
-
|
22 |
# Available models
|
23 |
AVAILABLE_MODELS = {
|
24 |
"distilgpt2": "distilgpt2",
|
@@ -28,58 +28,67 @@ AVAILABLE_MODELS = {
|
|
28 |
"pythia-160m": "EleutherAI/pythia-160m"
|
29 |
}
|
30 |
|
31 |
-
# Initialize model and tokenizer
|
32 |
current_model = None
|
33 |
current_tokenizer = None
|
34 |
current_model_name = None
|
|
|
35 |
|
36 |
def load_model(model_name):
|
|
|
37 |
global current_model, current_tokenizer, current_model_name
|
38 |
if current_model_name != model_name:
|
39 |
-
current_model = AutoModelForCausalLM.from_pretrained(AVAILABLE_MODELS[model_name])
|
40 |
current_tokenizer = AutoTokenizer.from_pretrained(AVAILABLE_MODELS[model_name])
|
41 |
current_model_name = model_name
|
42 |
|
|
|
|
|
|
|
43 |
def get_next_token_predictions(text, model_name, top_k=10):
|
|
|
44 |
global current_model, current_tokenizer
|
45 |
|
46 |
-
# Load model if
|
47 |
if current_model_name != model_name:
|
48 |
load_model(model_name)
|
49 |
|
50 |
-
|
51 |
-
|
52 |
with torch.no_grad():
|
53 |
outputs = current_model(**inputs)
|
54 |
logits = outputs.logits[0, -1, :]
|
55 |
probs = torch.nn.functional.softmax(logits, dim=-1)
|
56 |
-
|
57 |
top_k_probs, top_k_indices = torch.topk(probs, k=top_k)
|
58 |
top_k_tokens = [current_tokenizer.decode([idx.item()]) for idx in top_k_indices]
|
59 |
|
60 |
-
return top_k_tokens, top_k_probs.tolist()
|
61 |
|
62 |
-
def predict_next_token(model_name, text, custom_token=""):
|
63 |
-
|
64 |
if custom_token:
|
65 |
text += custom_token
|
66 |
-
|
67 |
-
|
68 |
-
|
69 |
-
|
70 |
-
# Format predictions
|
71 |
-
predictions = "\n".join([f"'{token}' : {prob:.4f}" for token, prob in zip(tokens, probs)])
|
72 |
|
73 |
return gr.update(choices=[f"'{t}'" for t in tokens]), predictions
|
74 |
|
75 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
76 |
with gr.Blocks() as demo:
|
77 |
-
gr.Markdown("# Interactive Text
|
78 |
-
|
79 |
-
|
80 |
-
|
81 |
-
|
82 |
-
""")
|
83 |
|
84 |
with gr.Row():
|
85 |
model_dropdown = gr.Dropdown(
|
@@ -91,31 +100,47 @@ with gr.Blocks() as demo:
|
|
91 |
with gr.Row():
|
92 |
text_input = gr.Textbox(
|
93 |
lines=5,
|
94 |
-
label="Text",
|
95 |
placeholder="Type your text here...",
|
96 |
value="The quick brown fox"
|
97 |
)
|
98 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
99 |
with gr.Row():
|
100 |
predict_button = gr.Button("Predict")
|
101 |
-
|
102 |
with gr.Row():
|
103 |
token_dropdown = gr.Dropdown(
|
104 |
-
label="Predicted
|
105 |
choices=[]
|
106 |
)
|
107 |
-
|
|
|
108 |
with gr.Row():
|
109 |
predictions_output = gr.Textbox(
|
110 |
lines=10,
|
111 |
-
label="Token
|
112 |
)
|
113 |
-
|
114 |
-
#
|
115 |
predict_button.click(
|
116 |
predict_next_token,
|
117 |
-
inputs=[model_dropdown, text_input],
|
118 |
outputs=[token_dropdown, predictions_output]
|
119 |
)
|
120 |
|
121 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
import os
|
2 |
+
import numpy as np
|
3 |
+
import torch
|
4 |
+
import gradio as gr
|
5 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer
|
6 |
+
|
7 |
+
# Handle Hugging Face Spaces GPU
|
8 |
if os.environ.get("SPACES_ZERO_GPU") is not None:
|
9 |
import spaces
|
10 |
else:
|
|
|
19 |
def fake_gpu():
|
20 |
pass
|
21 |
|
|
|
|
|
|
|
|
|
|
|
22 |
# Available models
|
23 |
AVAILABLE_MODELS = {
|
24 |
"distilgpt2": "distilgpt2",
|
|
|
28 |
"pythia-160m": "EleutherAI/pythia-160m"
|
29 |
}
|
30 |
|
31 |
+
# Initialize model and tokenizer
|
32 |
current_model = None
|
33 |
current_tokenizer = None
|
34 |
current_model_name = None
|
35 |
+
device = "cuda" if torch.cuda.is_available() else "cpu"
|
36 |
|
37 |
def load_model(model_name):
|
38 |
+
"""Load the selected model and tokenizer."""
|
39 |
global current_model, current_tokenizer, current_model_name
|
40 |
if current_model_name != model_name:
|
41 |
+
current_model = AutoModelForCausalLM.from_pretrained(AVAILABLE_MODELS[model_name]).to(device)
|
42 |
current_tokenizer = AutoTokenizer.from_pretrained(AVAILABLE_MODELS[model_name])
|
43 |
current_model_name = model_name
|
44 |
|
45 |
+
# Load the default model at startup
|
46 |
+
load_model("distilgpt2")
|
47 |
+
|
48 |
def get_next_token_predictions(text, model_name, top_k=10):
|
49 |
+
"""Generate the next token predictions with their probabilities."""
|
50 |
global current_model, current_tokenizer
|
51 |
|
52 |
+
# Load the model if it has changed
|
53 |
if current_model_name != model_name:
|
54 |
load_model(model_name)
|
55 |
|
56 |
+
inputs = current_tokenizer(text, return_tensors="pt").to(device)
|
57 |
+
|
58 |
with torch.no_grad():
|
59 |
outputs = current_model(**inputs)
|
60 |
logits = outputs.logits[0, -1, :]
|
61 |
probs = torch.nn.functional.softmax(logits, dim=-1)
|
62 |
+
|
63 |
top_k_probs, top_k_indices = torch.topk(probs, k=top_k)
|
64 |
top_k_tokens = [current_tokenizer.decode([idx.item()]) for idx in top_k_indices]
|
65 |
|
66 |
+
return top_k_tokens, top_k_probs.cpu().tolist()
|
67 |
|
68 |
+
def predict_next_token(model_name, text, top_k, custom_token=""):
|
69 |
+
"""Get predictions and update the UI."""
|
70 |
if custom_token:
|
71 |
text += custom_token
|
72 |
+
|
73 |
+
tokens, probs = get_next_token_predictions(text, model_name, top_k)
|
74 |
+
|
75 |
+
predictions = "\n".join([f"'{token}': {prob:.4f}" for token, prob in zip(tokens, probs)])
|
|
|
|
|
76 |
|
77 |
return gr.update(choices=[f"'{t}'" for t in tokens]), predictions
|
78 |
|
79 |
+
def append_selected_token(text, selected_token):
|
80 |
+
"""Append selected token from dropdown to the text input."""
|
81 |
+
if selected_token:
|
82 |
+
text += f" {selected_token.strip('\'')}"
|
83 |
+
return text
|
84 |
+
|
85 |
+
# Create the UI
|
86 |
with gr.Blocks() as demo:
|
87 |
+
gr.Markdown("# 🔥 Interactive Text Prediction with Transformers")
|
88 |
+
gr.Markdown(
|
89 |
+
"This application lets you interactively generate text using multiple transformer models. "
|
90 |
+
"Choose a model, type your text, and explore token predictions."
|
91 |
+
)
|
|
|
92 |
|
93 |
with gr.Row():
|
94 |
model_dropdown = gr.Dropdown(
|
|
|
100 |
with gr.Row():
|
101 |
text_input = gr.Textbox(
|
102 |
lines=5,
|
103 |
+
label="Input Text",
|
104 |
placeholder="Type your text here...",
|
105 |
value="The quick brown fox"
|
106 |
)
|
107 |
|
108 |
+
with gr.Row():
|
109 |
+
top_k_slider = gr.Slider(
|
110 |
+
minimum=1,
|
111 |
+
maximum=20,
|
112 |
+
value=10,
|
113 |
+
step=1,
|
114 |
+
label="Top-k Predictions"
|
115 |
+
)
|
116 |
+
|
117 |
with gr.Row():
|
118 |
predict_button = gr.Button("Predict")
|
119 |
+
|
120 |
with gr.Row():
|
121 |
token_dropdown = gr.Dropdown(
|
122 |
+
label="Predicted Tokens",
|
123 |
choices=[]
|
124 |
)
|
125 |
+
append_button = gr.Button("Append Token")
|
126 |
+
|
127 |
with gr.Row():
|
128 |
predictions_output = gr.Textbox(
|
129 |
lines=10,
|
130 |
+
label="Token Probabilities"
|
131 |
)
|
132 |
+
|
133 |
+
# Button click events
|
134 |
predict_button.click(
|
135 |
predict_next_token,
|
136 |
+
inputs=[model_dropdown, text_input, top_k_slider],
|
137 |
outputs=[token_dropdown, predictions_output]
|
138 |
)
|
139 |
|
140 |
+
append_button.click(
|
141 |
+
append_selected_token,
|
142 |
+
inputs=[text_input, token_dropdown],
|
143 |
+
outputs=text_input
|
144 |
+
)
|
145 |
+
|
146 |
+
demo.queue().launch()
|