Add llamacpp to models.py

modules/models.py

@@ -42,9 +42,10 @@ def load_model(model_name):
     t0 = time.time()
 
     shared.is_RWKV = 'rwkv-' in model_name.lower()
+    shared.is_llamacpp = model_name.lower().startswith('llamacpp-')
 
     # Default settings
-    if not any([shared.args.cpu, shared.args.load_in_8bit, shared.args.wbits, shared.args.auto_devices, shared.args.disk, shared.args.gpu_memory is not None, shared.args.cpu_memory is not None, shared.args.deepspeed, shared.args.flexgen, shared.is_RWKV]):
+    if not any([shared.args.cpu, shared.args.load_in_8bit, shared.args.wbits, shared.args.auto_devices, shared.args.disk, shared.args.gpu_memory is not None, shared.args.cpu_memory is not None, shared.args.deepspeed, shared.args.flexgen, shared.is_RWKV, shared.is_llamacpp]):
         if any(size in shared.model_name.lower() for size in ('13b', '20b', '30b')):
             model = AutoModelForCausalLM.from_pretrained(Path(f"{shared.args.model_dir}/{shared.model_name}"), device_map='auto', load_in_8bit=True)
         else:
@@ -100,6 +101,12 @@ def load_model(model_name):
 
         model = load_quantized(model_name)
 
+    # LLAMACPP model
+    elif shared.is_llamacpp:
+        from modules.llamacpp_model import LlamaCppModel
+        model, tokenizer = LlamaCppModel.from_pretrained(Path(f'models/{model_name}/ggml-model-q4_0.bin'))
+        return model, tokenizer
+
     # Custom
     else:
         params = {"low_cpu_mem_usage": True}