removing config.json and moving the setting to the app files

.dockerignore

@@ -7,7 +7,6 @@
 !Dockerfile
 !docker-compose.yml
 !*.sh
-!config.json
 
 # Include any other necessary files
 !model/**

_wsgi.py

@@ -1,5 +1,4 @@
 import os
-import json
 import logging
 import logging.config
 from pathlib import Path
@@ -35,25 +34,9 @@ logging.config.dictConfig({
 
 logger = logging.getLogger(__name__)
 
-def get_config():
-    """Load configuration from config.json"""
-    config_path = os.path.join(os.path.dirname(__file__), 'config.json')
-    if not os.path.exists(config_path):
-        logger.warning(f"Config file not found at {config_path}, using default settings")
-        return {}
-        
-    try:
-        with open(config_path) as f:
-            config = json.load(f)
-        return config
-    except Exception as e:
-        logger.error(f"Error loading config: {str(e)}")
-        return {}
-
 # Initialize the app at module level for Gunicorn
 app = init_app(
     model_class=T5Model,
-    config=get_config(),
     basic_auth_user=os.environ.get('BASIC_AUTH_USER'),
     basic_auth_pass=os.environ.get('BASIC_AUTH_PASS')
 )

config.json

@@ -1,20 +0,0 @@
-{
-    "model": {
-        "name": "google/flan-t5-base",
-        "max_length": 512,
-        "generation_max_length": 128,
-        "num_return_sequences": 1
-    },
-    "lora": {
-        "r": 8,
-        "alpha": 32,
-        "dropout": 0.1,
-        "target_modules": ["q", "v"]
-    },
-    "training": {
-        "learning_rate": 1e-4,
-        "batch_size": 1,
-        "max_steps": 100,
-        "save_steps": 50
-    }
-} 

docker-compose.yml

@@ -9,26 +9,29 @@ services:
       args:
         TEST_ENV: ${TEST_ENV}
     environment:
-      # specify these parameters if you want to use basic auth for the model server
+      # Model settings
+      - MODEL_NAME=google/flan-t5-base
+      - MAX_LENGTH=512
+      - GENERATION_MAX_LENGTH=128
+      - NUM_RETURN_SEQUENCES=1
+      # LoRA settings
+      - LORA_R=8
+      - LORA_ALPHA=32
+      - LORA_DROPOUT=0.1
+      - LORA_TARGET_MODULES=q,v
+      # Training settings
+      - LEARNING_RATE=1e-4
+      - BATCH_SIZE=1
+      - MAX_STEPS=100
+      - SAVE_STEPS=50
+      # Other settings
       - BASIC_AUTH_USER=
       - BASIC_AUTH_PASS=
-      # set the log level for the model server
       - LOG_LEVEL=DEBUG
-      # any other parameters that you want to pass to the model server
-      - ANY=PARAMETER
-      # specify the number of workers and threads for the model server
       - WORKERS=1
       - THREADS=8
-      # specify the model directory (likely you don't need to change this)
       - MODEL_DIR=/data/models
-
-      # Specify the Label Studio URL and API key to access
-      # uploaded, local storage and cloud storage files.
-      # Do not use 'localhost' as it does not work within Docker containers.
-      # Use prefix 'http://' or 'https://' for the URL always.
-      # Determine the actual IP using 'ifconfig' (Linux/Mac) or 'ipconfig' (Windows).
-      - LABEL_STUDIO_URL=
-      - LABEL_STUDIO_API_KEY=
+      - HF_CHECKPOINT_DIR=/data/checkpoints
     ports:
       - "9090:9090"
     volumes:

model.py

@@ -15,32 +15,26 @@ class T5Model(LabelStudioMLBase):
     def __init__(self, **kwargs):
         super(T5Model, self).__init__(**kwargs)
         
-        # Get configuration from kwargs (loaded from config.json if it exists)
-        config = kwargs.get('config', {})
-        model_config = config.get('model', {})
-        lora_config = config.get('lora', {})
-        training_config = config.get('training', {})
-        
-        # Model settings
-        self.model_name = model_config.get('name', "google/flan-t5-base")
-        self.max_length = model_config.get('max_length', 512)
-        self.generation_max_length = model_config.get('generation_max_length', 128)
-        self.num_return_sequences = model_config.get('num_return_sequences', 1)
+        # Model settings from environment variables
+        self.model_name = os.getenv('MODEL_NAME', 'google/flan-t5-base')
+        self.max_length = int(os.getenv('MAX_LENGTH', '512'))
+        self.generation_max_length = int(os.getenv('GENERATION_MAX_LENGTH', '128'))
+        self.num_return_sequences = int(os.getenv('NUM_RETURN_SEQUENCES', '1'))
         
         # LoRA settings
         self.lora_config = {
-            "r": lora_config.get('r', 8),
-            "alpha": lora_config.get('alpha', 32),
-            "dropout": lora_config.get('dropout', 0.1),
-            "target_modules": lora_config.get('target_modules', ["q", "v"])
+            "r": int(os.getenv('LORA_R', '8')),
+            "alpha": int(os.getenv('LORA_ALPHA', '32')),
+            "dropout": float(os.getenv('LORA_DROPOUT', '0.1')),
+            "target_modules": os.getenv('LORA_TARGET_MODULES', 'q,v').split(',')
         }
         
         # Training settings
         self.training_config = {
-            "learning_rate": training_config.get('learning_rate', 1e-4),
-            "batch_size": training_config.get('batch_size', 1),
-            "max_steps": training_config.get('max_steps', 100),
-            "save_steps": training_config.get('save_steps', 50)
+            "learning_rate": float(os.getenv('LEARNING_RATE', '1e-4')),
+            "batch_size": int(os.getenv('BATCH_SIZE', '1')),
+            "max_steps": int(os.getenv('MAX_STEPS', '100')),
+            "save_steps": int(os.getenv('SAVE_STEPS', '50'))
         }
         
         # Model components