New Models

Author: DarkStarStrix

.idea/encodings.xml

@@ -22,31 +22,38 @@
     python_requires=">=3.9",
     install_requires=[
         # Core ML
-        "torch>=2.4.1",
-        "pytorch-lightning>=2.4.0",
-        "transformers>=4.45.2",
-        "numpy>=1.24.4",
-        "pandas>=1.5.3",
+        "torch==2.1.1",
+        "pytorch-lightning==2.1.1",
+        "transformers==4.45.2",
+        "numpy==1.24.4",
+        "pandas==2.0.3",
+        "scipy==1.13.1",
 
         # API and Web
-        "fastapi>=0.109.1",
-        "uvicorn>=0.15.0",
-        "requests>=2.31.0",
-        "pydantic>=1.10.18",
-        "python-jose>=3.3.0",
-        "python-dotenv>=1.0.0",
+        "fastapi==0.109.1",
+        "uvicorn==0.15.0",
+        "requests==2.31.0",
+        "pydantic==1.10.18",
+        "python-jose==3.3.0",
+        "python-dotenv==1.0.0",
+        "PyJWT==2.10.0",
 
         # ML Tools
-        "scikit-learn>=1.3.2",
-        "seaborn>=0.11.2",
-        "matplotlib>=3.7.1",
-        "feature-engine>=1.0.0",
-        "optuna>=2.10.0",
-        "tensorboard>=2.12.0",
+        "scikit-learn==1.3.0",
+        "seaborn==0.11.2",
+        "matplotlib==3.7.1",
+        "optuna==4.1.0",
+        "config==0.5.1",
+        "plotly==5.24.1",
+
+        # Monitoring and Logging
+        "prometheus-client==0.21.0",
+        "rich==13.9.0",
+        "click==8.1.7",
     ],
     extras_require={
         "dev": [
-            "pytest>=7.4.3",
+            "pytest==7.4.3",
             "black>=22.0",
             "isort>=5.0",
             "flake8>=6.0.0",

Configs/setup.py

@@ -5,7 +5,7 @@
 from datetime import datetime
 import jwt
 import uuid
-from Config import (
+from Model_Library import (
     get_linear_config,
     get_logistic_regression_config,
     get_transformer_config,

Endpoints/app.py

@@ -269,6 +269,222 @@ def get_transformer_config():
     }
 
 
+def get_CNN_RNN_config():
+    """Configuration template for training basic CNN-RNN hybrid model."""
+    return {
+        "model": {
+            # CNN Parameters
+            "conv_layers": [32, 64, 128],  # Number of filters in each conv layer
+            "kernel_size": 3,
+            "pool_size": 2,
+            "cnn_activation": "relu",
+
+            # RNN Parameters
+            "rnn_type": "LSTM",  # LSTM or GRU
+            "hidden_size": 128,
+            "num_rnn_layers": 2,
+            "bidirectional": True,
+
+            # General Architecture
+            "input_channels": 1,
+            "dropout_rate": 0.5,
+            "final_dense_layers": [256, 128],
+            "output_size": 10  # Number of classes
+        },
+
+        "training": {
+            "num_epochs": 100,
+            "batch_size": 32,
+            "shuffle": True,
+            "validation_split": 0.2,
+            "early_stopping_patience": 10,
+            "save_best_only": True,
+            "monitor_metric": "val_accuracy"
+        },
+
+        "optimization": {
+            "optimizer": "adam",
+            "learning_rate": 0.001,
+            "weight_decay": 1e-4,
+            "lr_scheduler": "reduce_on_plateau",
+            "lr_patience": 5,
+            "lr_factor": 0.1,
+            "min_lr": 1e-6,
+            "clip_grad_norm": 1.0
+        },
+
+        "data": {
+            "dataset": "MNIST",
+            "data_dir": "./data",
+            "num_workers": 4,
+            "pin_memory": True,
+            "normalize": True,
+            "augmentation": {
+                "random_rotation": 10,
+                "random_zoom": 0.1,
+                "width_shift": 0.1,
+                "height_shift": 0.1
+            }
+        }
+    }
+
+
+def get_decision_tree_config():
+    return {
+        "model": {
+            "type": "decision_tree",
+            "input_dim": None,  # Set dynamically
+            "output_dim": None,  # Set dynamically
+            "max_depth": 5,
+            "min_samples_split": 2,
+            "min_samples_leaf": 1,
+            "max_features": "auto",
+            "criterion": "gini",
+            "task": "classification"
+        },
+        "training": {
+            "batch_size": 32,
+            "epochs": 10,
+            "validation_split": 0.2,
+            "early_stopping": True,
+            "patience": 3
+        },
+        "preprocessing": {
+            "scaling": None,  # Trees don't require scaling
+            "handle_missing": "median",
+            "handle_categorical": "label_encoding"
+        },
+        "logging": {
+            "tensorboard": True,
+            "log_interval": 10,
+            "metrics": ["accuracy", "precision", "recall", "f1"],
+            "feature_importance": True
+        }
+    }
+
+
+def get_naive_bayes_config():
+    return {
+        "model": {
+            "type": "naive_bayes",
+            "input_dim": None,  # Set dynamically based on data
+            "output_dim": None,  # Set dynamically based on data
+            "var_smoothing": 1e-9,  # Smoothing parameter
+            "priors": None,  # Class priors, None for automatic
+            "task": "classification"
+        },
+        "training": {
+            "batch_size": 32,
+            "epochs": 10,
+            "learning_rate": 0.001,
+            "early_stopping": True,
+            "patience": 5,
+            "validation_split": 0.2
+        },
+        "preprocessing": {
+            "scaling": "standard",
+            "handle_missing": "mean",
+            "feature_selection": None
+        },
+        "logging": {
+            "tensorboard": True,
+            "log_interval": 10,
+            "metrics": ["accuracy", "precision", "recall", "f1"]
+        }
+    }
+
+
+def get_Thermodynamic_Diffusion_config():
+    """Configuration template for training thermodynamic diffusion models."""
+    return {
+        "model": {
+            # Model Architecture
+            "image_size": 28,
+            "channels": 1,
+            "time_embedding_dim": 256,
+            "model_channels": 64,  # Base channel multiplier
+            "channel_multipliers": [1, 2, 4, 8],  # For different U-Net levels
+            "num_res_blocks": 2,  # Number of residual blocks per level
+            "attention_levels": [2, 3],  # Which levels to apply attention
+            "dropout_rate": 0.1,
+            "num_heads": 4,  # Number of attention heads
+
+            # Diffusion Process
+            "num_timesteps": 1000,
+            "beta_schedule": "linear",  # Options: linear, cosine, quadratic
+            "beta_start": 0.0001,
+            "beta_end": 0.02,
+
+            # Ising Model Parameters
+            "temperature_schedule": "linear",  # How temperature changes during diffusion
+            "initial_temperature": 0.1,
+            "final_temperature": 2.0,
+            "coupling_constant": 1.0,  # J in Ising model
+        },
+
+        "training": {
+            # Training Process
+            "num_epochs": 500,
+            "save_interval": 5000,
+            "eval_interval": 1000,
+            "log_interval": 100,
+            "sample_interval": 1000,
+            "num_samples": 64,  # Number of samples to generate during evaluation
+
+            # Loss Configuration
+            "loss_type": "l2",  # Options: l1, l2, huber
+            "loss_weight_type": "simple",  # Options: simple, snr, truncated
+
+            # Sampling Configuration
+            "sampling_steps": 250,  # Steps for fast sampling (< num_timesteps)
+            "clip_samples": True,
+            "clip_range": [-1, 1],
+        },
+
+        "optimization": {
+            # Optimizer Configuration
+            "optimizer": "AdamW",  # Options: Adam, AdamW, RMSprop
+            "learning_rate": 2e-4,
+            "weight_decay": 1e-4,
+            "eps": 1e-8,
+            "betas": (0.9, 0.999),
+
+            # Learning Rate Scheduling
+            "lr_schedule": "cosine",  # Options: cosine, linear, constant
+            "warmup_steps": 5000,
+            "min_lr": 1e-6,
+
+            # Gradient Configuration
+            "grad_clip": 1.0,
+            "ema_decay": 0.9999,  # Exponential moving average of model weights
+            "update_ema_interval": 1,
+        },
+
+        "data": {
+            # Data Configuration
+            "dataset": "MNIST",  # Dataset name
+            "data_dir": "./data",
+            "train_batch_size": 128,
+            "eval_batch_size": 256,
+
+            # Data Processing
+            "num_workers": 4,
+            "pin_memory": True,
+            "persistence": True,
+
+            # Augmentation
+            "random_flip": False,
+            "random_rotation": False,
+            "normalize": True,
+            "rescale": [-1, 1],  # Range to rescale images to
+
+            # Caching
+            "cache_size": 5000,  # Number of batches to cache in memory
+            "prefetch_factor": 2,
+        }
+    }
+
+
 def get_cnn_config():
     """Configuration for CNN models."""
     return {