Minor fixes to presets

app.py

@@ -80,8 +80,8 @@ def pecore(
         loaded_model = HuggingfaceModel.load(
             model_name_or_path,
             attribution_method,
-            model_kwargs=json.loads(model_kwargs),
-            tokenizer_kwargs=json.loads(tokenizer_kwargs),
+            model_kwargs={**json.loads(model_kwargs), **{"token": os.environ["HF_TOKEN"]}},
+            tokenizer_kwargs={**json.loads(tokenizer_kwargs), **{"token": os.environ["HF_TOKEN"]}},
         )
     if loaded_model.tokenizer.pad_token is None:
         loaded_model.tokenizer.add_special_tokens({"pad_token": "[PAD]"})
@@ -629,6 +629,7 @@ with gr.Blocks(css=custom_css) as demo:
         outputs=[
             model_name_or_path,
             input_template,
+            decoder_input_output_separator,
             contextless_input_template,
             special_tokens_to_keep,
             generation_kwargs,

presets.py

@@ -29,8 +29,9 @@ def set_default_preset():
 def set_zephyr_preset():
     return (
         "stabilityai/stablelm-2-zephyr-1_6b",  # model_name_or_path
-        "<|system|>{system_prompt}<|endoftext|>\n<|user|>\n{context}\n\n{current}<|endoftext|>\n<|assistant|>\n".replace("{system_prompt}", SYSTEM_PROMPT),  # input_template
-        "<|system|>{system_prompt}<|endoftext|>\n<|user|>\n{current}<|endoftext|>\n<|assistant|>\n".replace("{system_prompt}", SYSTEM_PROMPT),  # input_current_text_template
+        "<|system|>{system_prompt}<|endoftext|>\n<|user|>\n{context}\n\n{current}<|endoftext|>\n<|assistant|>".replace("{system_prompt}", SYSTEM_PROMPT),  # input_template
+        "\n",  # decoder_input_output_separator
+        "<|system|>{system_prompt}<|endoftext|>\n<|user|>\n{current}<|endoftext|>\n<|assistant|>".replace("{system_prompt}", SYSTEM_PROMPT),  # input_current_text_template
         ["<|im_start|>", "<|im_end|>", "<|endoftext|>"],  # special_tokens_to_keep
         '{\n\t"max_new_tokens": 50\n}',  # generation_kwargs
     )