Upload 3 files
Browse files- lstm_cnn_app.py +98 -0
- requirements.txt +7 -0
- tcn_app.py +107 -0
lstm_cnn_app.py
ADDED
@@ -0,0 +1,98 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
"""
|
2 |
+
lstm_cnn_app.py
|
3 |
+
Gradio app to serve the CNN-LSTM fault classification model.
|
4 |
+
|
5 |
+
Usage:
|
6 |
+
- Place a local model file named by LOCAL_MODEL_FILE in the same repo, or
|
7 |
+
- Set HUB_REPO and HUB_FILENAME to a public Hugging Face model repo + filename,
|
8 |
+
and the app will download it at startup using hf_hub_download.
|
9 |
+
"""
|
10 |
+
import os
|
11 |
+
import numpy as np
|
12 |
+
import pandas as pd
|
13 |
+
import gradio as gr
|
14 |
+
from tensorflow.keras.models import load_model
|
15 |
+
from huggingface_hub import hf_hub_download
|
16 |
+
|
17 |
+
# CONFIG: change these if your model filename/repo are different
|
18 |
+
LOCAL_MODEL_FILE = "lstm_cnn_model.h5"
|
19 |
+
HUB_REPO = "" # e.g., "username/lstm-cnn-model"
|
20 |
+
HUB_FILENAME = "" # e.g., "lstm_cnn_model.h5"
|
21 |
+
|
22 |
+
def get_model_path():
|
23 |
+
if os.path.exists(LOCAL_MODEL_FILE):
|
24 |
+
return LOCAL_MODEL_FILE
|
25 |
+
if HUB_REPO and HUB_FILENAME:
|
26 |
+
try:
|
27 |
+
print(f"Downloading {HUB_FILENAME} from {HUB_REPO} ...")
|
28 |
+
return hf_hub_download(repo_id=HUB_REPO, filename=HUB_FILENAME)
|
29 |
+
except Exception as e:
|
30 |
+
print("Failed to download from hub:", e)
|
31 |
+
return None
|
32 |
+
|
33 |
+
MODEL_PATH = get_model_path()
|
34 |
+
MODEL = None
|
35 |
+
if MODEL_PATH:
|
36 |
+
try:
|
37 |
+
MODEL = load_model(MODEL_PATH)
|
38 |
+
print("Loaded model:", MODEL_PATH)
|
39 |
+
except Exception as e:
|
40 |
+
print("Failed to load model:", e)
|
41 |
+
MODEL = None
|
42 |
+
else:
|
43 |
+
print("No model found. Please upload a model named", LOCAL_MODEL_FILE, "or set HUB_REPO/HUB_FILENAME.")
|
44 |
+
|
45 |
+
def prepare_input_array(arr, n_timesteps=1, n_features=None):
|
46 |
+
arr = np.array(arr)
|
47 |
+
if arr.ndim == 1:
|
48 |
+
if n_features is None:
|
49 |
+
return arr.reshape(1, n_timesteps, -1)
|
50 |
+
return arr.reshape(1, n_timesteps, n_features)
|
51 |
+
elif arr.ndim == 2:
|
52 |
+
return arr
|
53 |
+
else:
|
54 |
+
return arr
|
55 |
+
|
56 |
+
def predict_text(text, n_timesteps=1, n_features=None):
|
57 |
+
if MODEL is None:
|
58 |
+
return "模型未加载,请上传或配置模型。"
|
59 |
+
arr = np.fromstring(text, sep=',')
|
60 |
+
x = prepare_input_array(arr, n_timesteps=int(n_timesteps), n_features=(int(n_features) if n_features else None))
|
61 |
+
probs = MODEL.predict(x)
|
62 |
+
label = int(np.argmax(probs, axis=1)[0])
|
63 |
+
return f"预测类别: {label} (概率: {float(np.max(probs)):.4f})"
|
64 |
+
|
65 |
+
def predict_csv(file, n_timesteps=1, n_features=None):
|
66 |
+
if MODEL is None:
|
67 |
+
return {"error": "模型未加载,请上传或配置模型。"}
|
68 |
+
df = pd.read_csv(file.name)
|
69 |
+
X = df.values
|
70 |
+
if n_features:
|
71 |
+
X = X.reshape(X.shape[0], int(n_timesteps), int(n_features))
|
72 |
+
preds = MODEL.predict(X)
|
73 |
+
labels = preds.argmax(axis=1).tolist()
|
74 |
+
return {"labels": labels, "probs": preds.tolist()}
|
75 |
+
|
76 |
+
with gr.Blocks() as demo:
|
77 |
+
gr.Markdown("# CNN-LSTM Fault Classification")
|
78 |
+
gr.Markdown("上传 CSV(每行一个样本)或粘贴逗号分隔的一行特征进行预测。")
|
79 |
+
with gr.Row():
|
80 |
+
file_in = gr.File(label="上传 CSV(每行 = 一个样本)")
|
81 |
+
text_in = gr.Textbox(lines=2, placeholder="粘贴逗号分隔的一行特征,例如: 0.1,0.2,0.3,...")
|
82 |
+
n_ts = gr.Number(value=1, label="timesteps (整型)")
|
83 |
+
n_feat = gr.Number(value=None, label="features (可选,留空尝试自动推断)")
|
84 |
+
btn = gr.Button("预测")
|
85 |
+
out_text = gr.Textbox(label="单样本预测输出")
|
86 |
+
out_json = gr.JSON(label="批量预测结果 (labels & probs)")
|
87 |
+
|
88 |
+
def run_predict(file, text, n_timesteps, n_features):
|
89 |
+
if file is not None:
|
90 |
+
return "CSV 预测完成", predict_csv(file, n_timesteps, n_features)
|
91 |
+
if text:
|
92 |
+
return predict_text(text, n_timesteps, n_features), {}
|
93 |
+
return "请提供 CSV 或特征文本", {}
|
94 |
+
|
95 |
+
btn.click(run_predict, inputs=[file_in, text_in, n_ts, n_feat], outputs=[out_text, out_json])
|
96 |
+
|
97 |
+
if __name__ == '__main__':
|
98 |
+
demo.launch(server_name='0.0.0.0', server_port=7861)
|
requirements.txt
ADDED
@@ -0,0 +1,7 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
gradio>=3.0
|
2 |
+
tensorflow>=2.6
|
3 |
+
numpy
|
4 |
+
pandas
|
5 |
+
scikit-learn
|
6 |
+
huggingface_hub
|
7 |
+
matplotlib
|
tcn_app.py
ADDED
@@ -0,0 +1,107 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
"""
|
2 |
+
tcn_app.py
|
3 |
+
Gradio app to serve the TCN fault classification model.
|
4 |
+
|
5 |
+
Usage:
|
6 |
+
- Place a local model file named by LOCAL_MODEL_FILE in the same repo, or
|
7 |
+
- Set HUB_REPO and HUB_FILENAME to a public Hugging Face model repo + filename,
|
8 |
+
and the app will download it at startup using hf_hub_download.
|
9 |
+
|
10 |
+
This file is ready to push to a Hugging Face Space (Gradio).
|
11 |
+
"""
|
12 |
+
import os
|
13 |
+
import numpy as np
|
14 |
+
import pandas as pd
|
15 |
+
import gradio as gr
|
16 |
+
from tensorflow.keras.models import load_model
|
17 |
+
from huggingface_hub import hf_hub_download
|
18 |
+
|
19 |
+
# CONFIG: change these if your model filename/repo are different
|
20 |
+
LOCAL_MODEL_FILE = "tcn_model.h5"
|
21 |
+
HUB_REPO = "" # e.g., "username/tcn-model-repo" (leave empty to disable)
|
22 |
+
HUB_FILENAME = "" # e.g., "tcn_model.h5"
|
23 |
+
|
24 |
+
def get_model_path():
|
25 |
+
# prefer local
|
26 |
+
if os.path.exists(LOCAL_MODEL_FILE):
|
27 |
+
return LOCAL_MODEL_FILE
|
28 |
+
# try hub
|
29 |
+
if HUB_REPO and HUB_FILENAME:
|
30 |
+
try:
|
31 |
+
print(f"Downloading {HUB_FILENAME} from {HUB_REPO} ...")
|
32 |
+
return hf_hub_download(repo_id=HUB_REPO, filename=HUB_FILENAME)
|
33 |
+
except Exception as e:
|
34 |
+
print("Failed to download from hub:", e)
|
35 |
+
return None
|
36 |
+
|
37 |
+
MODEL_PATH = get_model_path()
|
38 |
+
MODEL = None
|
39 |
+
if MODEL_PATH:
|
40 |
+
try:
|
41 |
+
MODEL = load_model(MODEL_PATH)
|
42 |
+
print("Loaded model:", MODEL_PATH)
|
43 |
+
except Exception as e:
|
44 |
+
print("Failed to load model:", e)
|
45 |
+
MODEL = None
|
46 |
+
else:
|
47 |
+
print("No model found. Please upload a model named", LOCAL_MODEL_FILE, "or set HUB_REPO/HUB_FILENAME.")
|
48 |
+
|
49 |
+
def prepare_input_array(arr, n_timesteps=1, n_features=None):
|
50 |
+
arr = np.array(arr)
|
51 |
+
# If input is 1D, reshape to (1, n_timesteps, n_features)
|
52 |
+
if arr.ndim == 1:
|
53 |
+
if n_features is None:
|
54 |
+
# if user didn't supply n_features, assume arr is already shaped as (timesteps*features,)
|
55 |
+
return arr.reshape(1, n_timesteps, -1)
|
56 |
+
return arr.reshape(1, n_timesteps, n_features)
|
57 |
+
elif arr.ndim == 2:
|
58 |
+
# Already (timesteps, features) or (samples, features)
|
59 |
+
if arr.shape[0] == 1:
|
60 |
+
return arr.reshape(1, arr.shape[1], -1)
|
61 |
+
return arr
|
62 |
+
else:
|
63 |
+
return arr
|
64 |
+
|
65 |
+
def predict_text(text, n_timesteps=1, n_features=None):
|
66 |
+
if MODEL is None:
|
67 |
+
return "模型未加载,请上传或配置模型。"
|
68 |
+
arr = np.fromstring(text, sep=',')
|
69 |
+
x = prepare_input_array(arr, n_timesteps=int(n_timesteps), n_features=(int(n_features) if n_features else None))
|
70 |
+
probs = MODEL.predict(x)
|
71 |
+
label = int(np.argmax(probs, axis=1)[0])
|
72 |
+
return f"预测类别: {label} (概率: {float(np.max(probs)):.4f})"
|
73 |
+
|
74 |
+
def predict_csv(file, n_timesteps=1, n_features=None):
|
75 |
+
if MODEL is None:
|
76 |
+
return {"error": "模型未加载,请上传或配置模型。"}
|
77 |
+
df = pd.read_csv(file.name)
|
78 |
+
X = df.values
|
79 |
+
if n_features:
|
80 |
+
X = X.reshape(X.shape[0], int(n_timesteps), int(n_features))
|
81 |
+
preds = MODEL.predict(X)
|
82 |
+
labels = preds.argmax(axis=1).tolist()
|
83 |
+
return {"labels": labels, "probs": preds.tolist()}
|
84 |
+
|
85 |
+
with gr.Blocks() as demo:
|
86 |
+
gr.Markdown("# TCN Fault Classification")
|
87 |
+
gr.Markdown("上传 CSV(每行一个样本)或粘贴逗号分隔的一行特征进行预测。")
|
88 |
+
with gr.Row():
|
89 |
+
file_in = gr.File(label="上传 CSV(每行 = 一个样本)")
|
90 |
+
text_in = gr.Textbox(lines=2, placeholder="粘贴逗号分隔的一行特征,例如: 0.1,0.2,0.3,...")
|
91 |
+
n_ts = gr.Number(value=1, label="timesteps (整型)")
|
92 |
+
n_feat = gr.Number(value=None, label="features (可选,留空尝试自动推断)")
|
93 |
+
btn = gr.Button("预测")
|
94 |
+
out_text = gr.Textbox(label="单样本预测输出")
|
95 |
+
out_json = gr.JSON(label="批量预测结果 (labels & probs)")
|
96 |
+
|
97 |
+
def run_predict(file, text, n_timesteps, n_features):
|
98 |
+
if file is not None:
|
99 |
+
return "CSV 预测完成", predict_csv(file, n_timesteps, n_features)
|
100 |
+
if text:
|
101 |
+
return predict_text(text, n_timesteps, n_features), {}
|
102 |
+
return "请提供 CSV 或特征文本", {}
|
103 |
+
|
104 |
+
btn.click(run_predict, inputs=[file_in, text_in, n_ts, n_feat], outputs=[out_text, out_json])
|
105 |
+
|
106 |
+
if __name__ == '__main__':
|
107 |
+
demo.launch(server_name='0.0.0.0', server_port=7860)
|