lowvoltagenation commited on
Commit
ed31be4
·
1 Parent(s): 5457f70

Update model dropdown to show only Maya LoRA, Mistral v0.3, and Kimi K2

Browse files
Files changed (1) hide show
  1. src/model_interface.py +21 -47
src/model_interface.py CHANGED
@@ -43,58 +43,32 @@ class ModelInterface:
43
 
44
  logger.info(f"Model interface initialized on device: {self.device}")
45
 
46
- # Define available models
47
  self.available_models = {
48
- # Small/Fast models for quick testing
49
- "microsoft/DialoGPT-small": {
50
- "name": "DialoGPT Small",
51
- "description": "Fast conversational model for testing",
52
- "size": "Small (~300MB)",
53
- "type": "local"
54
- },
55
- "microsoft/DialoGPT-medium": {
56
- "name": "DialoGPT Medium",
57
- "description": "Balanced conversational model",
58
- "size": "Medium (~1GB)",
59
- "type": "local"
60
- },
61
-
62
- # Larger models (will use quantization)
63
- "meta-llama/Llama-2-7b-chat-hf": {
64
- "name": "Llama 2 7B Chat",
65
- "description": "Meta's Llama 2 optimized for chat",
66
- "size": "Large (~7B params)",
67
- "type": "local",
68
  "requires_auth": True
69
  },
70
- "mistralai/Mistral-7B-Instruct-v0.1": {
71
- "name": "Mistral 7B Instruct",
72
- "description": "Mistral's instruction-tuned model",
 
73
  "size": "Large (~7B params)",
74
- "type": "local"
75
- },
76
-
77
- # HF Inference API models (no local loading required)
78
- "gpt2": {
79
- "name": "GPT-2",
80
- "description": "OpenAI's GPT-2 (via HF Inference API)",
81
- "size": "API",
82
- "type": "inference_api"
83
- },
84
- "microsoft/DialoGPT-large": {
85
- "name": "DialoGPT Large",
86
- "description": "Large conversational model (via API)",
87
- "size": "API",
88
- "type": "inference_api"
89
  },
90
-
91
- # Placeholder for fine-tuned models
92
- "blakeurmos/maya-finetuned-v1": {
93
- "name": "Maya Fine-tuned v1",
94
- "description": "Custom fine-tuned Maya model",
95
- "size": "Custom",
96
- "type": "custom",
97
- "exists": False # Will check if exists
98
  }
99
  }
100
 
 
43
 
44
  logger.info(f"Model interface initialized on device: {self.device}")
45
 
46
+ # Define available models (optimized for HuggingFace Spaces)
47
  self.available_models = {
48
+ # Maya's fine-tuned LoRA model
49
+ "blakeurmos/maya-7b-lora-v1": {
50
+ "name": "Maya 7B (Fine-tuned)",
51
+ "description": "Maya's personality fine-tuned on Mistral-7B",
52
+ "size": "LoRA (~14MB + base model)",
53
+ "type": "lora",
54
+ "base_model": "mistralai/Mistral-7B-Instruct-v0.3",
 
 
 
 
 
 
 
 
 
 
 
 
 
55
  "requires_auth": True
56
  },
57
+ # Latest Mistral instruction model
58
+ "mistralai/Mistral-7B-Instruct-v0.3": {
59
+ "name": "Mistral 7B Instruct v0.3",
60
+ "description": "Mistral's latest instruction-tuned model",
61
  "size": "Large (~7B params)",
62
+ "type": "inference_api",
63
+ "requires_auth": True
 
 
 
 
 
 
 
 
 
 
 
 
 
64
  },
65
+ # Moonshot AI's latest model
66
+ "moonshotai/Kimi-K2-Instruct": {
67
+ "name": "Kimi K2 Instruct",
68
+ "description": "Moonshot AI's latest instruction model",
69
+ "size": "Large",
70
+ "type": "inference_api",
71
+ "requires_auth": True
 
72
  }
73
  }
74