IBM / unitxt / 15870906710 / 1

81%

main: 81%

push

github

Add multi turn tool calling task and support multiple tools per call (#1811)

* Add multi turn tool calling task and tools

Signed-off-by: elronbandel <elronbandel@gmail.com>

* remove unrelated changes

Signed-off-by: elronbandel <elronbandel@gmail.com>

* Update all the single turn single call datasets in bfcl

Signed-off-by: elronbandel <elronbandel@gmail.com>

* Add llm as judge for tool calling

Signed-off-by: elronbandel <elronbandel@gmail.com>

* Add some more dataset

Signed-off-by: elronbandel <elronbandel@gmail.com>

* Update prepare/tasks/tool_calling/tasks.py

Co-authored-by: Yoav Katz <68273864+yoavkatz@users.noreply.github.com>

* Some fixes

Signed-off-by: elronbandel <elronbandel@gmail.com>

* Revert execute expression mistake

Signed-off-by: elronbandel <elronbandel@gmail.com>

* Add xlam function calling

Signed-off-by: elronbandel <elronbandel@gmail.com>

* Add tool calling benchmark

Signed-off-by: elronbandel <elronbandel@gmail.com>

* Fix typing

Signed-off-by: elronbandel <elronbandel@gmail.com>

* Enhanced multi turn support and examples

Signed-off-by: elronbandel <elronbandel@gmail.com>

* Fix serializers test

Signed-off-by: elronbandel <elronbandel@gmail.com>

* Fix example

Signed-off-by: elronbandel <elronbandel@gmail.com>

---------

Signed-off-by: elronbandel <elronbandel@gmail.com>
Co-authored-by: Yoav Katz <68273864+yoavkatz@users.noreply.github.com>

1698 of 2104 branches covered (80.7%)

Branch coverage included in aggregate %.

10563 of 13244 relevant lines covered (79.76%)

0.8 hits per line