Update formats (#769)

updated formats

Co-authored-by: michal <[email protected]>
Co-authored-by: Yoav Katz <[email protected]>

Author: 3 people

prepare/formats/empty.py

@@ -2,8 +2,8 @@
 from unitxt.formats import SystemFormat
 
 format = SystemFormat(
-    demo_format="{source}\n {target_prefix}{target}\n\n",
-    model_input_format="{system_prompt}{instruction}{demos}{source}\n{target_prefix}",
+    demo_format="{source}\\N{target_prefix}{target}\n\n",
+    model_input_format="{system_prompt}\\N{instruction}\\N{demos}{source}\\N{target_prefix}",
 )
 
 add_to_catalog(format, "formats.empty", overwrite=True)

prepare/formats/models/llama2.py

@@ -9,7 +9,7 @@
 # {{ user_msg_1 }} [/INST] {{ model_answer_1 }} </s><s>[INST] {{ user_msg_2 }} [/INST]
 format = SystemFormat(
     demo_format="{source} [/INST] {target_prefix}{target} </s><s>[INST] ",
-    model_input_format="[INST] <<SYS>>\n{system_prompt}\n\n{instruction}<<SYS>>\n\n\n{demos}{source} [/INST]{target_prefix}",
+    model_input_format="[INST] <<SYS>>\n{system_prompt}\\N{instruction}<</SYS>>\n\n\n{demos}{source} [/INST] {target_prefix}",
 )
 
 add_to_catalog(format, "formats.llama2", overwrite=True)

src/unitxt/catalog/formats/empty.json

@@ -1,5 +1,5 @@
 {
     "type": "system_format",
-    "demo_format": "{source}\n {target_prefix}{target}\n\n",
-    "model_input_format": "{system_prompt}{instruction}{demos}{source}\n{target_prefix}"
+    "demo_format": "{source}\\N{target_prefix}{target}\n\n",
+    "model_input_format": "{system_prompt}\\N{instruction}\\N{demos}{source}\\N{target_prefix}"
 }

src/unitxt/catalog/formats/llama2.json

@@ -1,5 +1,5 @@
 {
     "type": "system_format",
     "demo_format": "{source} [/INST] {target_prefix}{target} </s><s>[INST] ",
-    "model_input_format": "[INST] <<SYS>>\n{system_prompt}\n\n{instruction}<<SYS>>\n\n\n{demos}{source} [/INST]{target_prefix}"
+    "model_input_format": "[INST] <<SYS>>\n{system_prompt}\\N{instruction}<</SYS>>\n\n\n{demos}{source} [/INST] {target_prefix}"
 }