File size: 6,081 Bytes
20d720d
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
"""
Models module for Personal Coach CrewAI Application
Handles all AI model loading and management
"""

from typing import TYPE_CHECKING, Optional, Dict, Any
import torch

# Version info
__version__ = "1.0.0"

# Lazy imports
if TYPE_CHECKING:
    from .mistral_model import MistralModel, MistralConfig, MistralPromptFormatter

# Public API
__all__ = [
    # Main model classes
    "MistralModel",
    "MistralConfig", 
    "MistralPromptFormatter",
    
    # Model management
    "load_model",
    "get_model_info",
    "clear_model_cache",
    
    # Constants
    "AVAILABLE_MODELS",
    "MODEL_REQUIREMENTS",
    "DEFAULT_MODEL_CONFIG"
]

# Available models
AVAILABLE_MODELS = {
    "mistral-7b-instruct": {
        "model_id": "mistralai/Mistral-7B-Instruct-v0.1",
        "type": "instruction-following",
        "size": "7B",
        "context_length": 32768,
        "languages": ["multilingual"]
    },
    "mistral-7b": {
        "model_id": "mistralai/Mistral-7B-v0.1",
        "type": "base",
        "size": "7B", 
        "context_length": 32768,
        "languages": ["multilingual"]
    }
}

# Model requirements
MODEL_REQUIREMENTS = {
    "mistral-7b-instruct": {
        "ram": "16GB",
        "vram": "8GB (GPU) or 16GB (CPU)",
        "disk": "15GB",
        "compute": "GPU recommended"
    }
}

# Default configuration
DEFAULT_MODEL_CONFIG = {
    "max_length": 2048,
    "temperature": 0.7,
    "top_p": 0.95,
    "top_k": 50,
    "do_sample": True,
    "num_return_sequences": 1,
    "device": "cuda" if torch.cuda.is_available() else "cpu",
    "torch_dtype": torch.float16 if torch.cuda.is_available() else torch.float32,
    "load_in_8bit": False,
    "cache_dir": ".cache/models"
}

# Model instance cache
_model_cache: Dict[str, Any] = {}

def load_model(model_name: str = "mistral-7b-instruct", config: Optional[Dict[str, Any]] = None):
    """
    Load a model with caching support
    
    Args:
        model_name: Name of the model to load
        config: Optional configuration override
        
    Returns:
        Model instance
    """
    # Check cache first
    cache_key = f"{model_name}_{str(config)}"
    if cache_key in _model_cache:
        return _model_cache[cache_key]
    
    # Import here to avoid circular imports
    from .mistral_model import MistralModel, MistralConfig
    
    # Get model info
    model_info = AVAILABLE_MODELS.get(model_name)
    if not model_info:
        raise ValueError(f"Unknown model: {model_name}")
    
    # Merge configurations
    model_config = DEFAULT_MODEL_CONFIG.copy()
    if config:
        model_config.update(config)
    
    # Create config object
    mistral_config = MistralConfig(
        model_id=model_info["model_id"],
        **model_config
    )
    
    # Load model
    model = MistralModel(mistral_config)
    
    # Cache it
    _model_cache[cache_key] = model
    
    return model

def get_model_info(model_name: str) -> Optional[Dict[str, Any]]:
    """
    Get information about a model
    
    Args:
        model_name: Name of the model
        
    Returns:
        Model information dictionary or None
    """
    info = AVAILABLE_MODELS.get(model_name)
    if info:
        # Add requirements
        requirements = MODEL_REQUIREMENTS.get(model_name, {})
        info["requirements"] = requirements
        
        # Add loading status
        cache_keys = [k for k in _model_cache.keys() if k.startswith(model_name)]
        info["is_loaded"] = len(cache_keys) > 0
        
    return info

def clear_model_cache(model_name: Optional[str] = None):
    """
    Clear model cache to free memory
    
    Args:
        model_name: Specific model to clear, or None for all
    """
    global _model_cache
    
    if model_name:
        # Clear specific model
        keys_to_remove = [k for k in _model_cache.keys() if k.startswith(model_name)]
        for key in keys_to_remove:
            del _model_cache[key]
    else:
        # Clear all
        _model_cache.clear()
    
    # Force garbage collection
    import gc
    gc.collect()
    
    # Clear GPU cache if using CUDA
    if torch.cuda.is_available():
        torch.cuda.empty_cache()

# Utility functions
def estimate_memory_usage(model_name: str) -> Dict[str, Any]:
    """
    Estimate memory usage for a model
    
    Args:
        model_name: Name of the model
        
    Returns:
        Memory estimation dictionary
    """
    model_info = AVAILABLE_MODELS.get(model_name)
    if not model_info:
        return {}
    
    size = model_info.get("size", "7B")
    size_gb = float(size.replace("B", ""))
    
    estimates = {
        "model_size_gb": size_gb,
        "fp32_memory_gb": size_gb * 4,  # 4 bytes per parameter
        "fp16_memory_gb": size_gb * 2,  # 2 bytes per parameter
        "int8_memory_gb": size_gb,      # 1 byte per parameter
        "recommended_ram_gb": size_gb * 2.5,
        "recommended_vram_gb": size_gb * 1.5
    }
    
    return estimates

def get_device_info() -> Dict[str, Any]:
    """Get information about available compute devices"""
    info = {
        "cuda_available": torch.cuda.is_available(),
        "device_count": torch.cuda.device_count() if torch.cuda.is_available() else 0,
        "current_device": torch.cuda.current_device() if torch.cuda.is_available() else None,
        "device_name": torch.cuda.get_device_name() if torch.cuda.is_available() else "CPU"
    }
    
    if torch.cuda.is_available():
        info["gpu_memory"] = {
            "allocated": torch.cuda.memory_allocated() / 1024**3,  # GB
            "reserved": torch.cuda.memory_reserved() / 1024**3,    # GB
            "total": torch.cuda.get_device_properties(0).total_memory / 1024**3  # GB
        }
    
    return info

# Module initialization
import os
if os.getenv("DEBUG_MODE", "false").lower() == "true":
    print(f"Models module v{__version__} initialized")
    device_info = get_device_info()
    print(f"Device: {device_info['device_name']}")
    if device_info['cuda_available']:
        print(f"GPU Memory: {device_info['gpu_memory']['total']:.1f}GB")