Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
@@ -2,7 +2,6 @@ import torch
|
|
2 |
import gradio as gr
|
3 |
import multiprocessing
|
4 |
import os
|
5 |
-
import time
|
6 |
from transformers import AutoModelForCausalLM, AutoTokenizer, TrainingArguments, Trainer
|
7 |
from peft import get_peft_model, LoraConfig, TaskType
|
8 |
from datasets import load_dataset
|
@@ -11,34 +10,56 @@ device = "cpu"
|
|
11 |
training_process = None
|
12 |
log_file = "training_status.log"
|
13 |
|
|
|
14 |
def log_status(message):
|
15 |
with open(log_file, "w") as f:
|
16 |
f.write(message)
|
17 |
|
|
|
18 |
def read_status():
|
19 |
if os.path.exists(log_file):
|
20 |
with open(log_file, "r") as f:
|
21 |
return f.read()
|
22 |
return "⏳ در انتظار شروع ترینینگ..."
|
23 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
24 |
def train_model(dataset_url, model_url, epochs):
|
25 |
try:
|
26 |
log_status("🚀 در حال بارگیری مدل...")
|
|
|
27 |
tokenizer = AutoTokenizer.from_pretrained(model_url, trust_remote_code=True)
|
28 |
model = AutoModelForCausalLM.from_pretrained(
|
29 |
model_url, trust_remote_code=True, torch_dtype=torch.float32, device_map="cpu"
|
30 |
)
|
31 |
|
32 |
lora_config = LoraConfig(
|
33 |
-
task_type=TaskType.CAUSAL_LM,
|
|
|
|
|
|
|
|
|
34 |
)
|
35 |
-
|
36 |
model = get_peft_model(model, lora_config)
|
37 |
model.to(device)
|
38 |
|
39 |
dataset = load_dataset(dataset_url)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
40 |
def tokenize_function(examples):
|
41 |
-
return tokenizer(examples[
|
42 |
|
43 |
tokenized_datasets = dataset.map(tokenize_function, batched=True)
|
44 |
train_dataset = tokenized_datasets["train"]
|
@@ -61,8 +82,8 @@ def train_model(dataset_url, model_url, epochs):
|
|
61 |
)
|
62 |
|
63 |
trainer = Trainer(
|
64 |
-
model=model,
|
65 |
-
args=training_args,
|
66 |
train_dataset=train_dataset
|
67 |
)
|
68 |
|
@@ -78,6 +99,7 @@ def train_model(dataset_url, model_url, epochs):
|
|
78 |
except Exception as e:
|
79 |
log_status(f"❌ خطا: {str(e)}")
|
80 |
|
|
|
81 |
def start_training(dataset_url, model_url, epochs):
|
82 |
global training_process
|
83 |
if training_process is None or not training_process.is_alive():
|
@@ -87,26 +109,24 @@ def start_training(dataset_url, model_url, epochs):
|
|
87 |
else:
|
88 |
return "⚠ ترینینگ در حال اجرا است!"
|
89 |
|
|
|
90 |
def update_status():
|
91 |
return read_status()
|
92 |
|
|
|
93 |
with gr.Blocks() as app:
|
94 |
gr.Markdown("# 🚀 AutoTrain DeepSeek R1 (CPU) - نمایش وضعیت لحظهای")
|
95 |
|
96 |
-
|
97 |
-
|
98 |
-
|
99 |
-
|
100 |
-
train_button = gr.Button("شروع ترینینگ")
|
101 |
-
output_text = gr.Textbox(label="وضعیت ترینینگ")
|
102 |
-
|
103 |
-
train_button.click(start_training, inputs=[dataset_url, model_url, epochs], outputs=output_text)
|
104 |
|
105 |
-
|
106 |
-
|
107 |
-
refresh_button = gr.Button("🔄 بهروزرسانی وضعیت")
|
108 |
|
109 |
-
|
|
|
|
|
110 |
|
111 |
-
app.
|
112 |
-
app.launch(server_name="0.0.0.0", server_port=7860, share=True)
|
|
|
2 |
import gradio as gr
|
3 |
import multiprocessing
|
4 |
import os
|
|
|
5 |
from transformers import AutoModelForCausalLM, AutoTokenizer, TrainingArguments, Trainer
|
6 |
from peft import get_peft_model, LoraConfig, TaskType
|
7 |
from datasets import load_dataset
|
|
|
10 |
training_process = None
|
11 |
log_file = "training_status.log"
|
12 |
|
13 |
+
# Logging function
|
14 |
def log_status(message):
|
15 |
with open(log_file, "w") as f:
|
16 |
f.write(message)
|
17 |
|
18 |
+
# Read training status
|
19 |
def read_status():
|
20 |
if os.path.exists(log_file):
|
21 |
with open(log_file, "r") as f:
|
22 |
return f.read()
|
23 |
return "⏳ در انتظار شروع ترینینگ..."
|
24 |
|
25 |
+
# Function to find the text column dynamically
|
26 |
+
def find_text_column(dataset):
|
27 |
+
sample = dataset["train"][0] # Get the first row of the training dataset
|
28 |
+
for column in sample.keys():
|
29 |
+
if isinstance(sample[column], str): # Find the first text-like column
|
30 |
+
return column
|
31 |
+
return None # No valid text column found
|
32 |
+
|
33 |
+
# Model training function
|
34 |
def train_model(dataset_url, model_url, epochs):
|
35 |
try:
|
36 |
log_status("🚀 در حال بارگیری مدل...")
|
37 |
+
|
38 |
tokenizer = AutoTokenizer.from_pretrained(model_url, trust_remote_code=True)
|
39 |
model = AutoModelForCausalLM.from_pretrained(
|
40 |
model_url, trust_remote_code=True, torch_dtype=torch.float32, device_map="cpu"
|
41 |
)
|
42 |
|
43 |
lora_config = LoraConfig(
|
44 |
+
task_type=TaskType.CAUSAL_LM,
|
45 |
+
r=8,
|
46 |
+
lora_alpha=32,
|
47 |
+
lora_dropout=0.1,
|
48 |
+
target_modules=["q_proj", "v_proj"]
|
49 |
)
|
|
|
50 |
model = get_peft_model(model, lora_config)
|
51 |
model.to(device)
|
52 |
|
53 |
dataset = load_dataset(dataset_url)
|
54 |
+
|
55 |
+
# Automatically detect the correct text column
|
56 |
+
text_column = find_text_column(dataset)
|
57 |
+
if not text_column:
|
58 |
+
log_status("❌ خطا: ستون متنی در دیتاست یافت نشد!")
|
59 |
+
return
|
60 |
+
|
61 |
def tokenize_function(examples):
|
62 |
+
return tokenizer(examples[text_column], truncation=True, padding="max_length", max_length=256)
|
63 |
|
64 |
tokenized_datasets = dataset.map(tokenize_function, batched=True)
|
65 |
train_dataset = tokenized_datasets["train"]
|
|
|
82 |
)
|
83 |
|
84 |
trainer = Trainer(
|
85 |
+
model=model,
|
86 |
+
args=training_args,
|
87 |
train_dataset=train_dataset
|
88 |
)
|
89 |
|
|
|
99 |
except Exception as e:
|
100 |
log_status(f"❌ خطا: {str(e)}")
|
101 |
|
102 |
+
# Start training in a separate process
|
103 |
def start_training(dataset_url, model_url, epochs):
|
104 |
global training_process
|
105 |
if training_process is None or not training_process.is_alive():
|
|
|
109 |
else:
|
110 |
return "⚠ ترینینگ در حال اجرا است!"
|
111 |
|
112 |
+
# Function to update the status
|
113 |
def update_status():
|
114 |
return read_status()
|
115 |
|
116 |
+
# Gradio UI
|
117 |
with gr.Blocks() as app:
|
118 |
gr.Markdown("# 🚀 AutoTrain DeepSeek R1 (CPU) - نمایش وضعیت لحظهای")
|
119 |
|
120 |
+
with gr.Row():
|
121 |
+
dataset_input = gr.Textbox(label="📂 لینک دیتاست (Hugging Face)")
|
122 |
+
model_input = gr.Textbox(label="🤖 مدل پایه (Hugging Face)")
|
123 |
+
epochs_input = gr.Number(label="🔄 تعداد Epochs", value=3)
|
|
|
|
|
|
|
|
|
124 |
|
125 |
+
start_button = gr.Button("🚀 شروع ترینینگ")
|
126 |
+
status_output = gr.Textbox(label="📢 وضعیت ترینینگ", interactive=False)
|
|
|
127 |
|
128 |
+
start_button.click(start_training, inputs=[dataset_input, model_input, epochs_input], outputs=status_output)
|
129 |
+
status_button = gr.Button("🔄 بروزرسانی وضعیت")
|
130 |
+
status_button.click(update_status, outputs=status_output)
|
131 |
|
132 |
+
app.launch()
|
|