A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 16, 07:00 UTC
Change granite chat template to keep json list formatting for tool calls (#10452)
Signed-off-by: Max de Bayser <maxdebayser@gmail.com>
Maximilien de Bayser committed
Nov 19, 2024 at 22:16 UTC
a324d3a1a74ab0a3fafc0f2d19860bd1d1301a85
1 file changed
+1
-5
examples/tool_chat_template_granite.jinja
+1
-5
@@ -21,11 +21,7 @@
21
{{- '<|start_of_role|>user<|end_of_role|>' + message['content'] + '<|end_of_text|>
22
' }}
23
{%- elif message['role'] == 'assistant_tool_call' or (message['role'] == 'assistant' and message.tool_calls is defined) %}
24
- {{- '<|start_of_role|>assistant<|end_of_role|>' }}
25
- {% for tc in message.tool_calls %}
26
- {{- '<|tool_call|> ' + {'name': tc.function.name, 'arguments': tc.function.arguments}|tojson }}
27
- {% endfor %}
28
- {{- '<|end_of_text|>
24
+ {{- '<|start_of_role|>assistant<|end_of_role|><|tool_call|>' + message.tool_calls|map(attribute='function')|list|tojson(indent=4) + '<|end_of_text|>
25
' }}
26
{%- elif message['role'] == 'assistant' %}
27
{{- '<|start_of_role|>assistant<|end_of_role|>' + message['content'] + '<|end_of_text|>