onnx issue fixed

Author: niralbhalodia

multimind/__init__.py

@@ -243,9 +243,7 @@ def __init__(self, *args, **kwargs):
     BaseModelConverter,
     HuggingFaceConverter,
     OllamaConverter,
-    ONNXConverter,
     TensorFlowConverter,
-    ONNXRuntimeConverter,
     SafetensorsConverter,
     GGMLConverter,
     OptimizationConverter,
@@ -257,6 +255,17 @@ def __init__(self, *args, **kwargs):
     ModelConversionManager
 )
 
+# Try to import ONNX-related converters, but handle gracefully if not available
+try:
+    from .model_conversion import ONNXConverter
+except ImportError:
+    ONNXConverter = None
+
+try:
+    from .model_conversion import ONNXRuntimeConverter
+except ImportError:
+    ONNXRuntimeConverter = None
+
 # Context window components
 from .context_window import (
     ContextManager,

multimind/model_conversion/__init__.py

@@ -10,10 +10,25 @@
 # Core converters
 from .huggingface import HuggingFaceConverter
 from .ollama import OllamaConverter
-from .onnx import ONNXConverter
+
+# Try to import ONNXConverter, but handle gracefully if not available
+try:
+    from .onnx import ONNXConverter
+    ONNX_CONVERTER_AVAILABLE = True
+except ImportError:
+    ONNX_CONVERTER_AVAILABLE = False
+    ONNXConverter = None
 
 # Format converters
-from .formats import TensorFlowConverter, ONNXRuntimeConverter, SafetensorsConverter, GGMLConverter
+from .formats import TensorFlowConverter, SafetensorsConverter, GGMLConverter
+
+# Try to import ONNXRuntimeConverter, but handle gracefully if not available
+try:
+    from .formats import ONNXRuntimeConverter
+    ONNX_RUNTIME_CONVERTER_AVAILABLE = True
+except ImportError:
+    ONNX_RUNTIME_CONVERTER_AVAILABLE = False
+    ONNXRuntimeConverter = None
 
 # Optimization converters
 from .optimization import OptimizationConverter, AdvancedOptimization
@@ -34,11 +49,15 @@
     # Core converters
     'HuggingFaceConverter',
     'OllamaConverter',
-    'ONNXConverter',
-    
+]
+
+# Conditionally add ONNX-related exports
+if ONNX_CONVERTER_AVAILABLE:
+    __all__.append('ONNXConverter')
+
+__all__.extend([
     # Format converters
     'TensorFlowConverter',
-    'ONNXRuntimeConverter',
     'SafetensorsConverter',
     'GGMLConverter',
 
@@ -58,4 +77,8 @@
 
     # Manager
     'ModelConversionManager',
-] 
+])
+
+# Conditionally add ONNXRuntimeConverter
+if ONNX_RUNTIME_CONVERTER_AVAILABLE:
+    __all__.append('ONNXRuntimeConverter') 

multimind/model_conversion/formats.py

@@ -1,8 +1,6 @@
 from typing import Dict, Any, Optional
 from pathlib import Path
 import torch
-import onnx
-import onnxruntime
 from .base import BaseModelConverter
 
 # Try to import tensorflow, but handle gracefully if not available
@@ -13,6 +11,16 @@
     TENSORFLOW_AVAILABLE = False
     tf = None
 
+# Try to import onnx and onnxruntime, but handle gracefully if not available
+try:
+    import onnx
+    import onnxruntime
+    ONNX_AVAILABLE = True
+except ImportError:
+    ONNX_AVAILABLE = False
+    onnx = None
+    onnxruntime = None
+
 class TensorFlowConverter(BaseModelConverter):
     """Converter for TensorFlow models."""
 
@@ -85,6 +93,9 @@ def convert(self,
                 output_path: str,
                 config: Optional[Dict[str, Any]] = None) -> str:
         """Convert ONNX model to optimized ONNX Runtime format."""
+        if not ONNX_AVAILABLE:
+            raise ImportError("ONNX is not available. Please install onnx and onnxruntime to use this converter.")
+        
         config = config or {}
 
         # Load ONNX model
@@ -104,6 +115,8 @@ def _optimize_model(self, model: onnx.ModelProto, config: Dict[str, Any]) -> onn
 
     def validate(self, model_path: str) -> bool:
         """Validate ONNX model."""
+        if not ONNX_AVAILABLE:
+            return False
         try:
             onnx.load(model_path)
             return True
@@ -112,6 +125,9 @@ def validate(self, model_path: str) -> bool:
 
     def get_metadata(self, model_path: str) -> Dict[str, Any]:
         """Get ONNX model metadata."""
+        if not ONNX_AVAILABLE:
+            return {"format": "onnx", "error": "ONNX not available"}
+        
         model = onnx.load(model_path)
         return {
             "format": "onnx",

multimind/model_conversion/onnx.py

@@ -1,10 +1,17 @@
 import os
-import onnx
 import torch
 from typing import Dict, Any, Optional
 from transformers import AutoModelForCausalLM, AutoTokenizer
 from .base import BaseModelConverter
 
+# Try to import onnx, but handle gracefully if not available
+try:
+    import onnx
+    ONNX_AVAILABLE = True
+except ImportError:
+    ONNX_AVAILABLE = False
+    onnx = None
+
 class ONNXConverter(BaseModelConverter):
     """Converter for ONNX models."""
 
@@ -33,6 +40,9 @@ def convert(self,
         Returns:
             str: Path to the converted model
         """
+        if not ONNX_AVAILABLE:
+            raise ImportError("ONNX is not available. Please install onnx to use this converter.")
+        
         if not self.validate(model_path):
             raise ValueError(f"Invalid model path: {model_path}")
 
@@ -118,6 +128,9 @@ def get_metadata(self, model_path: str) -> Dict[str, Any]:
         Returns:
             Dict[str, Any]: Model metadata
         """
+        if not ONNX_AVAILABLE:
+            return {"error": "ONNX is not available"}
+        
         model = AutoModelForCausalLM.from_pretrained(model_path)
         tokenizer = AutoTokenizer.from_pretrained(model_path)