suppport for alpaca-like instruction datasets without inputs

scripts/finetune.py

@@ -102,8 +102,8 @@ def load_model(base_model, base_model_config, model_type, tokenizer_type, cfg, a
                 base_model_config if base_model_config else base_model,
                 model_path,
                 device_map=cfg.device_map,
-                groupsize=-1,
-                is_v1_model=True,
+                groupsize=cfg.gptq_groupsize if cfg.gptq_groupsize else -1,
+                is_v1_model=cfg.gptq_model_v1 if cfg.gptq_model_v1 is not None else True,
             )
             load_in_8bit = False
         elif "llama" in base_model:

src/axolotl/prompt_tokenizers.py

@@ -37,7 +37,8 @@ class AlpacaPromptTokenizingStrategy(PromptTokenizingStrategy):
         tokenized_full_prompt = self._tokenize(full_prompt)
         if not self.train_on_inputs:
             user_prompt = self.prompter.build_prompt(
-                prompt["instruction"], prompt["input"]
+                prompt["instruction"],
+                prompt["input"] if "input" in prompt else "",
             )
             tokenized_user_prompt = self._tokenize(user_prompt, add_eos_token=False)
             user_prompt_len = len(tokenized_user_prompt["input_ids"])
@@ -51,7 +52,7 @@ class AlpacaPromptTokenizingStrategy(PromptTokenizingStrategy):
     def _tokenize_full_prompt(self, prompt):
         return self.prompter.build_prompt(
             prompt["instruction"],
-            prompt["input"],
+            prompt["input"] if "input" in prompt else "",
             prompt["output"],
         )