Spaces:
Sleeping
Sleeping
Upload app.py with huggingface_hub
Browse files
app.py
CHANGED
|
@@ -22,8 +22,13 @@ logger = logging.getLogger(__name__)
|
|
| 22 |
# Load environment variables
|
| 23 |
load_dotenv()
|
| 24 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 25 |
# Load config file
|
| 26 |
def load_config(config_path="transformers_config.json"):
|
|
|
|
| 27 |
try:
|
| 28 |
with open(config_path, 'r') as f:
|
| 29 |
config = json.load(f)
|
|
@@ -43,8 +48,22 @@ SPACE_NAME = os.getenv("HF_SPACE_NAME", "phi4training")
|
|
| 43 |
# Function to run training in a thread and stream output to container logs
|
| 44 |
def run_training():
|
| 45 |
"""Run the training script and stream its output to container logs"""
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 46 |
process = subprocess.Popen(
|
| 47 |
-
[
|
| 48 |
stdout=subprocess.PIPE,
|
| 49 |
stderr=subprocess.STDOUT,
|
| 50 |
universal_newlines=True,
|
|
@@ -62,6 +81,8 @@ def start_training():
|
|
| 62 |
# Print directly to container logs
|
| 63 |
print("\n===== STARTING TRAINING PROCESS =====\n")
|
| 64 |
print(f"Model: {MODEL_NAME}")
|
|
|
|
|
|
|
| 65 |
print(f"Training with configuration from transformers_config.json")
|
| 66 |
print("Training logs will appear below:")
|
| 67 |
print("=" * 50)
|
|
@@ -147,7 +168,12 @@ with gr.Blocks(css="footer {visibility: hidden}") as demo:
|
|
| 147 |
if __name__ == "__main__":
|
| 148 |
# Start Gradio with minimal features
|
| 149 |
print("\n===== RESEARCH TRAINING DASHBOARD STARTED =====\n")
|
| 150 |
-
print("
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 151 |
print("All training output will appear in these logs")
|
| 152 |
logger.info("Starting research training dashboard")
|
| 153 |
demo.launch(share=False)
|
|
|
|
| 22 |
# Load environment variables
|
| 23 |
load_dotenv()
|
| 24 |
|
| 25 |
+
# Get script directory - important for Hugging Face Space paths
|
| 26 |
+
SCRIPT_DIR = os.path.dirname(os.path.abspath(__file__))
|
| 27 |
+
BASE_DIR = os.path.abspath(os.path.join(SCRIPT_DIR, "."))
|
| 28 |
+
|
| 29 |
# Load config file
|
| 30 |
def load_config(config_path="transformers_config.json"):
|
| 31 |
+
config_path = os.path.join(BASE_DIR, config_path)
|
| 32 |
try:
|
| 33 |
with open(config_path, 'r') as f:
|
| 34 |
config = json.load(f)
|
|
|
|
| 48 |
# Function to run training in a thread and stream output to container logs
|
| 49 |
def run_training():
|
| 50 |
"""Run the training script and stream its output to container logs"""
|
| 51 |
+
# Locate training script using absolute path
|
| 52 |
+
training_script = os.path.join(BASE_DIR, "run_cloud_training.py")
|
| 53 |
+
|
| 54 |
+
# Check if file exists and log the path
|
| 55 |
+
if not os.path.exists(training_script):
|
| 56 |
+
print(f"ERROR: Training script not found at: {training_script}")
|
| 57 |
+
print(f"Current directory: {os.getcwd()}")
|
| 58 |
+
print("Available files:")
|
| 59 |
+
for file in os.listdir(BASE_DIR):
|
| 60 |
+
print(f" - {file}")
|
| 61 |
+
return
|
| 62 |
+
|
| 63 |
+
print(f"Found training script at: {training_script}")
|
| 64 |
+
|
| 65 |
process = subprocess.Popen(
|
| 66 |
+
[sys.executable, training_script],
|
| 67 |
stdout=subprocess.PIPE,
|
| 68 |
stderr=subprocess.STDOUT,
|
| 69 |
universal_newlines=True,
|
|
|
|
| 81 |
# Print directly to container logs
|
| 82 |
print("\n===== STARTING TRAINING PROCESS =====\n")
|
| 83 |
print(f"Model: {MODEL_NAME}")
|
| 84 |
+
print(f"Base directory: {BASE_DIR}")
|
| 85 |
+
print(f"Current working directory: {os.getcwd()}")
|
| 86 |
print(f"Training with configuration from transformers_config.json")
|
| 87 |
print("Training logs will appear below:")
|
| 88 |
print("=" * 50)
|
|
|
|
| 168 |
if __name__ == "__main__":
|
| 169 |
# Start Gradio with minimal features
|
| 170 |
print("\n===== RESEARCH TRAINING DASHBOARD STARTED =====\n")
|
| 171 |
+
print(f"Base directory: {BASE_DIR}")
|
| 172 |
+
print(f"Current working directory: {os.getcwd()}")
|
| 173 |
+
print("Available files:")
|
| 174 |
+
for file in os.listdir(BASE_DIR):
|
| 175 |
+
print(f" - {file}")
|
| 176 |
+
print("\nClick 'Start Training' to begin the fine-tuning process")
|
| 177 |
print("All training output will appear in these logs")
|
| 178 |
logger.info("Starting research training dashboard")
|
| 179 |
demo.launch(share=False)
|