Lazy import + refactor Lora layer addition (#426)

* lazy model import in mlx_lm

* change lora loading

* fix olmo lora

* remove a bunch of unused stuff from plamo

* move phixtral to mlx-lm and out of llms/

llms/mlx_lm/models/olmo.py

@@ -7,18 +7,25 @@ import mlx.nn as nn
 
 from .base import BaseModelArgs
 
+try:
+    import hf_olmo
+except ImportError:
+    print("To run olmo install ai2-olmo: pip install ai2-olmo")
+    exit(1)
+
 
 @dataclass
 class ModelArgs(BaseModelArgs):
+    model_type: str
     d_model: int
     n_layers: int
     mlp_hidden_size: int
     n_heads: int
     vocab_size: int
     embedding_size: int
+    model_type: str
     rope_theta: float = 10000
     rope_traditional: bool = False
-    model_type: str = None
     mlp_ratio: int = 4
     weight_tying: bool = False
 
@@ -162,11 +169,7 @@ class OlmoModel(nn.Module):
 class Model(nn.Module):
     def __init__(self, args: ModelArgs):
         super().__init__()
-        try:
-            import hf_olmo
-        except ImportError:
-            print("To run olmo install ai2-olmo: pip install ai2-olmo")
-            exit(1)
+        self.model_type = args.model_type
         self.model = OlmoModel(args)
 
     def __call__(