Spaces:
Running
Running
Sadjad Alikhani
commited on
Update app.py
Browse files
app.py
CHANGED
@@ -3,6 +3,8 @@ import os
|
|
3 |
from PIL import Image
|
4 |
import numpy as np
|
5 |
import pickle
|
|
|
|
|
6 |
|
7 |
# Paths to the predefined images folder
|
8 |
RAW_PATH = os.path.join("images", "raw")
|
@@ -13,21 +15,29 @@ GENERATED_PATH = os.path.join("images", "generated")
|
|
13 |
percentage_values = [10, 30, 50, 70, 100]
|
14 |
complexity_values = [16, 32]
|
15 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
16 |
# Function to load and display predefined images based on user selection
|
17 |
def display_predefined_images(percentage_idx, complexity_idx):
|
18 |
-
# Map the slider index to the actual value
|
19 |
percentage = percentage_values[percentage_idx]
|
20 |
complexity = complexity_values[complexity_idx]
|
21 |
-
|
22 |
-
# Generate the paths to the images
|
23 |
raw_image_path = os.path.join(RAW_PATH, f"percentage_{percentage}_complexity_{complexity}.png")
|
24 |
embeddings_image_path = os.path.join(EMBEDDINGS_PATH, f"percentage_{percentage}_complexity_{complexity}.png")
|
25 |
|
26 |
-
# Load images using PIL
|
27 |
raw_image = Image.open(raw_image_path)
|
28 |
embeddings_image = Image.open(embeddings_image_path)
|
29 |
|
30 |
-
# Return the loaded images
|
31 |
return raw_image, embeddings_image
|
32 |
|
33 |
import torch
|
@@ -35,16 +45,17 @@ import subprocess
|
|
35 |
|
36 |
# Function to load the pre-trained model from your cloned repository
|
37 |
def load_custom_model():
|
38 |
-
# Assume your model is in the cloned LWM repository
|
39 |
from lwm_model import LWM # Assuming the model is defined in lwm_model.py
|
40 |
model = LWM() # Modify this according to your model initialization
|
41 |
-
model.eval()
|
42 |
return model
|
43 |
|
44 |
# Function to process the uploaded .p file and perform inference using the custom model
|
45 |
def process_p_file(uploaded_file, percentage_idx, complexity_idx):
|
|
|
|
|
|
|
46 |
try:
|
47 |
-
# Clone the repository if not already done (for model and tokenizer)
|
48 |
model_repo_url = "https://huggingface.co/sadjadalikhani/LWM"
|
49 |
model_repo_dir = "./LWM"
|
50 |
|
@@ -52,49 +63,45 @@ def process_p_file(uploaded_file, percentage_idx, complexity_idx):
|
|
52 |
print(f"Cloning model repository from {model_repo_url}...")
|
53 |
subprocess.run(["git", "clone", model_repo_url, model_repo_dir], check=True)
|
54 |
|
55 |
-
# Change the working directory to the cloned LWM folder
|
56 |
if os.path.exists(model_repo_dir):
|
57 |
os.chdir(model_repo_dir)
|
58 |
print(f"Changed working directory to {os.getcwd()}")
|
59 |
else:
|
60 |
return f"Directory {model_repo_dir} does not exist."
|
61 |
|
62 |
-
# Step 1: Load the custom model
|
63 |
from lwm_model import LWM
|
64 |
device = 'cuda' if torch.cuda.is_available() else 'cpu'
|
65 |
print(f"Loading the LWM model on {device}...")
|
66 |
model = LWM.from_pretrained(device=device)
|
67 |
|
68 |
-
# Step 2: Import the tokenizer
|
69 |
from input_preprocess import tokenizer
|
70 |
|
71 |
-
# Step 3: Load the uploaded .p file that contains the wireless channel matrix
|
72 |
with open(uploaded_file.name, 'rb') as f:
|
73 |
manual_data = pickle.load(f)
|
74 |
|
75 |
-
# Step 4: Tokenize the data if needed (or perform any necessary preprocessing)
|
76 |
preprocessed_chs = tokenizer(manual_data=manual_data)
|
77 |
|
78 |
-
# Step 5: Perform inference on the channel matrix using the model
|
79 |
from inference import lwm_inference, create_raw_dataset
|
80 |
output_emb = lwm_inference(preprocessed_chs, 'channel_emb', model)
|
81 |
output_raw = create_raw_dataset(preprocessed_chs, device)
|
82 |
-
print(output_emb.shape)
|
83 |
-
print(output_raw.shape)
|
84 |
|
85 |
-
|
|
|
|
|
|
|
86 |
|
87 |
except Exception as e:
|
88 |
-
return str(e), str(e)
|
|
|
|
|
|
|
89 |
|
90 |
# Function to handle logic based on whether a file is uploaded or not
|
91 |
def los_nlos_classification(file, percentage_idx, complexity_idx):
|
92 |
if file is not None:
|
93 |
-
# Process the uploaded .p file and generate new images
|
94 |
return process_p_file(file, percentage_idx, complexity_idx)
|
95 |
else:
|
96 |
-
|
97 |
-
return display_predefined_images(percentage_idx, complexity_idx)
|
98 |
|
99 |
# Define the Gradio interface
|
100 |
with gr.Blocks(css="""
|
@@ -126,7 +133,6 @@ with gr.Blocks(css="""
|
|
126 |
with gr.Tab("Beam Prediction Task"):
|
127 |
gr.Markdown("### Beam Prediction Task")
|
128 |
|
129 |
-
# Sliders for percentage and complexity
|
130 |
with gr.Row():
|
131 |
with gr.Column(elem_id="slider-container"):
|
132 |
gr.Markdown("Percentage of Data for Training")
|
@@ -135,22 +141,18 @@ with gr.Blocks(css="""
|
|
135 |
gr.Markdown("Task Complexity")
|
136 |
complexity_slider_bp = gr.Slider(minimum=0, maximum=1, step=1, value=0, interactive=True, elem_id="vertical-slider")
|
137 |
|
138 |
-
# Image outputs (display the images side by side and set a smaller size for the images)
|
139 |
with gr.Row():
|
140 |
raw_img_bp = gr.Image(label="Raw Channels", type="pil", width=300, height=300, interactive=False)
|
141 |
embeddings_img_bp = gr.Image(label="Embeddings", type="pil", width=300, height=300, interactive=False)
|
142 |
|
143 |
-
# Instant image updates when sliders change
|
144 |
percentage_slider_bp.change(fn=display_predefined_images, inputs=[percentage_slider_bp, complexity_slider_bp], outputs=[raw_img_bp, embeddings_img_bp])
|
145 |
complexity_slider_bp.change(fn=display_predefined_images, inputs=[percentage_slider_bp, complexity_slider_bp], outputs=[raw_img_bp, embeddings_img_bp])
|
146 |
|
147 |
with gr.Tab("LoS/NLoS Classification Task"):
|
148 |
gr.Markdown("### LoS/NLoS Classification Task")
|
149 |
|
150 |
-
# File uploader for uploading .p file
|
151 |
file_input = gr.File(label="Upload .p File", file_types=[".p"])
|
152 |
|
153 |
-
# Sliders for percentage and complexity
|
154 |
with gr.Row():
|
155 |
with gr.Column(elem_id="slider-container"):
|
156 |
gr.Markdown("Percentage of Data for Training")
|
@@ -159,15 +161,14 @@ with gr.Blocks(css="""
|
|
159 |
gr.Markdown("Task Complexity")
|
160 |
complexity_slider_los = gr.Slider(minimum=0, maximum=1, step=1, value=0, interactive=True, elem_id="vertical-slider")
|
161 |
|
162 |
-
# Image outputs (display the images side by side and set a smaller size for the images)
|
163 |
with gr.Row():
|
164 |
raw_img_los = gr.Image(label="Raw Channels", type="pil", width=300, height=300, interactive=False)
|
165 |
embeddings_img_los = gr.Image(label="Embeddings", type="pil", width=300, height=300, interactive=False)
|
|
|
166 |
|
167 |
-
|
168 |
-
|
169 |
-
|
170 |
-
complexity_slider_los.change(fn=los_nlos_classification, inputs=[file_input, percentage_slider_los, complexity_slider_los], outputs=[raw_img_los, embeddings_img_los])
|
171 |
|
172 |
# Launch the app
|
173 |
if __name__ == "__main__":
|
|
|
3 |
from PIL import Image
|
4 |
import numpy as np
|
5 |
import pickle
|
6 |
+
import io
|
7 |
+
import sys
|
8 |
|
9 |
# Paths to the predefined images folder
|
10 |
RAW_PATH = os.path.join("images", "raw")
|
|
|
15 |
percentage_values = [10, 30, 50, 70, 100]
|
16 |
complexity_values = [16, 32]
|
17 |
|
18 |
+
# Custom class to capture print output
|
19 |
+
class PrintCapture(io.StringIO):
|
20 |
+
def __init__(self):
|
21 |
+
super().__init__()
|
22 |
+
self.output = []
|
23 |
+
|
24 |
+
def write(self, txt):
|
25 |
+
self.output.append(txt)
|
26 |
+
super().write(txt)
|
27 |
+
|
28 |
+
def get_output(self):
|
29 |
+
return ''.join(self.output)
|
30 |
+
|
31 |
# Function to load and display predefined images based on user selection
|
32 |
def display_predefined_images(percentage_idx, complexity_idx):
|
|
|
33 |
percentage = percentage_values[percentage_idx]
|
34 |
complexity = complexity_values[complexity_idx]
|
|
|
|
|
35 |
raw_image_path = os.path.join(RAW_PATH, f"percentage_{percentage}_complexity_{complexity}.png")
|
36 |
embeddings_image_path = os.path.join(EMBEDDINGS_PATH, f"percentage_{percentage}_complexity_{complexity}.png")
|
37 |
|
|
|
38 |
raw_image = Image.open(raw_image_path)
|
39 |
embeddings_image = Image.open(embeddings_image_path)
|
40 |
|
|
|
41 |
return raw_image, embeddings_image
|
42 |
|
43 |
import torch
|
|
|
45 |
|
46 |
# Function to load the pre-trained model from your cloned repository
|
47 |
def load_custom_model():
|
|
|
48 |
from lwm_model import LWM # Assuming the model is defined in lwm_model.py
|
49 |
model = LWM() # Modify this according to your model initialization
|
50 |
+
model.eval()
|
51 |
return model
|
52 |
|
53 |
# Function to process the uploaded .p file and perform inference using the custom model
|
54 |
def process_p_file(uploaded_file, percentage_idx, complexity_idx):
|
55 |
+
capture = PrintCapture()
|
56 |
+
sys.stdout = capture # Redirect print statements to capture
|
57 |
+
|
58 |
try:
|
|
|
59 |
model_repo_url = "https://huggingface.co/sadjadalikhani/LWM"
|
60 |
model_repo_dir = "./LWM"
|
61 |
|
|
|
63 |
print(f"Cloning model repository from {model_repo_url}...")
|
64 |
subprocess.run(["git", "clone", model_repo_url, model_repo_dir], check=True)
|
65 |
|
|
|
66 |
if os.path.exists(model_repo_dir):
|
67 |
os.chdir(model_repo_dir)
|
68 |
print(f"Changed working directory to {os.getcwd()}")
|
69 |
else:
|
70 |
return f"Directory {model_repo_dir} does not exist."
|
71 |
|
|
|
72 |
from lwm_model import LWM
|
73 |
device = 'cuda' if torch.cuda.is_available() else 'cpu'
|
74 |
print(f"Loading the LWM model on {device}...")
|
75 |
model = LWM.from_pretrained(device=device)
|
76 |
|
|
|
77 |
from input_preprocess import tokenizer
|
78 |
|
|
|
79 |
with open(uploaded_file.name, 'rb') as f:
|
80 |
manual_data = pickle.load(f)
|
81 |
|
|
|
82 |
preprocessed_chs = tokenizer(manual_data=manual_data)
|
83 |
|
|
|
84 |
from inference import lwm_inference, create_raw_dataset
|
85 |
output_emb = lwm_inference(preprocessed_chs, 'channel_emb', model)
|
86 |
output_raw = create_raw_dataset(preprocessed_chs, device)
|
|
|
|
|
87 |
|
88 |
+
print(f"Output Embeddings Shape: {output_emb.shape}")
|
89 |
+
print(f"Output Raw Shape: {output_raw.shape}")
|
90 |
+
|
91 |
+
return output_emb, output_raw, capture.get_output()
|
92 |
|
93 |
except Exception as e:
|
94 |
+
return str(e), str(e), capture.get_output()
|
95 |
+
|
96 |
+
finally:
|
97 |
+
sys.stdout = sys.__stdout__ # Reset print statements
|
98 |
|
99 |
# Function to handle logic based on whether a file is uploaded or not
|
100 |
def los_nlos_classification(file, percentage_idx, complexity_idx):
|
101 |
if file is not None:
|
|
|
102 |
return process_p_file(file, percentage_idx, complexity_idx)
|
103 |
else:
|
104 |
+
return display_predefined_images(percentage_idx, complexity_idx), None
|
|
|
105 |
|
106 |
# Define the Gradio interface
|
107 |
with gr.Blocks(css="""
|
|
|
133 |
with gr.Tab("Beam Prediction Task"):
|
134 |
gr.Markdown("### Beam Prediction Task")
|
135 |
|
|
|
136 |
with gr.Row():
|
137 |
with gr.Column(elem_id="slider-container"):
|
138 |
gr.Markdown("Percentage of Data for Training")
|
|
|
141 |
gr.Markdown("Task Complexity")
|
142 |
complexity_slider_bp = gr.Slider(minimum=0, maximum=1, step=1, value=0, interactive=True, elem_id="vertical-slider")
|
143 |
|
|
|
144 |
with gr.Row():
|
145 |
raw_img_bp = gr.Image(label="Raw Channels", type="pil", width=300, height=300, interactive=False)
|
146 |
embeddings_img_bp = gr.Image(label="Embeddings", type="pil", width=300, height=300, interactive=False)
|
147 |
|
|
|
148 |
percentage_slider_bp.change(fn=display_predefined_images, inputs=[percentage_slider_bp, complexity_slider_bp], outputs=[raw_img_bp, embeddings_img_bp])
|
149 |
complexity_slider_bp.change(fn=display_predefined_images, inputs=[percentage_slider_bp, complexity_slider_bp], outputs=[raw_img_bp, embeddings_img_bp])
|
150 |
|
151 |
with gr.Tab("LoS/NLoS Classification Task"):
|
152 |
gr.Markdown("### LoS/NLoS Classification Task")
|
153 |
|
|
|
154 |
file_input = gr.File(label="Upload .p File", file_types=[".p"])
|
155 |
|
|
|
156 |
with gr.Row():
|
157 |
with gr.Column(elem_id="slider-container"):
|
158 |
gr.Markdown("Percentage of Data for Training")
|
|
|
161 |
gr.Markdown("Task Complexity")
|
162 |
complexity_slider_los = gr.Slider(minimum=0, maximum=1, step=1, value=0, interactive=True, elem_id="vertical-slider")
|
163 |
|
|
|
164 |
with gr.Row():
|
165 |
raw_img_los = gr.Image(label="Raw Channels", type="pil", width=300, height=300, interactive=False)
|
166 |
embeddings_img_los = gr.Image(label="Embeddings", type="pil", width=300, height=300, interactive=False)
|
167 |
+
output_textbox = gr.Textbox(label="Console Output", lines=10)
|
168 |
|
169 |
+
file_input.change(fn=los_nlos_classification, inputs=[file_input, percentage_slider_los, complexity_slider_los], outputs=[raw_img_los, embeddings_img_los, output_textbox])
|
170 |
+
percentage_slider_los.change(fn=los_nlos_classification, inputs=[file_input, percentage_slider_los, complexity_slider_los], outputs=[raw_img_los, embeddings_img_los, output_textbox])
|
171 |
+
complexity_slider_los.change(fn=los_nlos_classification, inputs=[file_input, percentage_slider_los, complexity_slider_los], outputs=[raw_img_los, embeddings_img_los, output_textbox])
|
|
|
172 |
|
173 |
# Launch the app
|
174 |
if __name__ == "__main__":
|