quickfix, bad PR happened

Author: cartazio

pyproject.toml

@@ -16,7 +16,7 @@ dependencies = [
     "watchdog>=3.0.0",
     "xdg>=6.0.0",
     "accelerate>=0.21.0",
-    "transformers>=4.18.0",
+    "transformers>=4.32.1",
     "flask>=2.3.2",
     "flask_cors>=4.0.0",
     "llama-index>=0.7.7",

src/llm_vm/onsite_llm.py

@@ -10,8 +10,6 @@
     BertTokenizer,
     OPTForCausalLM,
     BloomForCausalLM,
-    Llama2Tokenizer,
-    Llama2ForCausalLM,
     GPTNeoForCausalLM,
     GPTNeoXForCausalLM,
     GPT2Tokenizer,
@@ -300,9 +298,9 @@ class Small_Local_LLama(Base_Onsite_LLM):
     model_uri="meta-llama/Llama-2-7b"
 
     def model_loader(self):
-        return Llama2ForCausalLM.from_pretrained(self.model_uri)
+        return LlamaForCausalLM.from_pretrained(self.model_uri)
     def tokenizer_loader(self):
-        return Llama2Tokenizer.from_pretrained(self.model_uri)
+        return AutoTokenizer.from_pretrained(self.model_uri)
 
 @RegisterModelClass("flan")# our yummiest model based on similarity to food
 class Small_Local_Flan_T5(Base_Onsite_LLM):