Merge remote-tracking branch 'refs/remotes/origin/main'

modules/LoRA.py

@@ -114,11 +114,12 @@ def add_lora_transformers(lora_names):
     if len(lora_names) > 0:
         params = {}
         if not shared.args.cpu:
-            params['dtype'] = shared.model.dtype
-            if hasattr(shared.model, "hf_device_map"):
-                params['device_map'] = {"base_model.model." + k: v for k, v in shared.model.hf_device_map.items()}
-            elif shared.args.load_in_8bit:
-                params['device_map'] = {'': 0}
+            if shared.args.load_in_4bit or shared.args.load_in_8bit: 
+                params['peft_type'] = shared.model.dtype
+            else:
+                params['dtype'] = shared.model.dtype
+                if hasattr(shared.model, "hf_device_map"):
+                    params['device_map'] = {"base_model.model." + k: v for k, v in shared.model.hf_device_map.items()}
 
         logger.info("Applying the following LoRAs to {}: {}".format(shared.model_name, ', '.join(lora_names)))
         shared.model = PeftModel.from_pretrained(shared.model, Path(f"{shared.args.lora_dir}/{lora_names[0]}"), adapter_name=lora_names[0], **params)

modules/models.py

@@ -95,11 +95,18 @@ def load_tokenizer(model_name, model):
     if any(s in model_name.lower() for s in ['gpt-4chan', 'gpt4chan']) and Path(f"{shared.args.model_dir}/gpt-j-6B/").exists():
         tokenizer = AutoTokenizer.from_pretrained(Path(f"{shared.args.model_dir}/gpt-j-6B/"))
     elif path_to_model.exists():
-        tokenizer = AutoTokenizer.from_pretrained(
-            path_to_model,
-            trust_remote_code=shared.args.trust_remote_code,
-            use_fast=False
-        )
+        try:
+            tokenizer = AutoTokenizer.from_pretrained(
+                path_to_model,
+                trust_remote_code=shared.args.trust_remote_code,
+                use_fast=False
+            )
+        except ValueError:
+             tokenizer = AutoTokenizer.from_pretrained(
+                path_to_model,
+                trust_remote_code=shared.args.trust_remote_code,
+                use_fast=True
+            )           
 
     if tokenizer.__class__.__name__ == 'LlamaTokenizer':
         pairs = [

training/formats/vicuna-format.json

@@ -0,0 +1,3 @@
+{
+  "instruction,output": "A chat between a curious user and an artificial intelligence assistant. The assistant gives helpful, detailed, and polite answers to the user's questions.\n\nUSER: %instruction%\n\nASSISTANT: %output%"
+}