Allow tool calls through together-py and remove alternating roles check

[connermanuel]

This PR makes changes to the files validation logic to allow tool calling datasets through. To do this:

• we allow tools to exist alongside messages for conversational data
• we allow messages to have "tool_calls" instead of "content"
• we allow the role of "tool"
• we remove the requirement for alternating roles (since tools can be called sequentially).

Tested by uploading a relevant file.

---

Note

Medium Risk
Validation rules are loosened for conversational datasets (role alternation and content requirements), which could allow previously-rejected malformed training data to pass if the new constraints are insufficient.

Overview
Updates JSONL dataset validation to support tool-calling conversations by allowing a top-level tools field, permitting message roles to include tool, and accepting messages that contain tool_calls instead of content.

Removes the strict user/assistant alternation rule and adjusts message required fields accordingly, with unit tests updated to drop the alternation assertions and to reflect the new required-field behavior.

^{Written by Cursor Bugbot for commit 984e182. This will update automatically on new commits. Configure here.}

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{Bugbot Autofix is OFF. To automatically fix reported issues with Cloud Agents, enable Autofix in the Cursor dashboard.}

This PR is being reviewed by Cursor Bugbot

Details

Your team is on the Bugbot Free tier. On this plan, Bugbot will review limited PRs each billing cycle for each member of your team.

To receive Bugbot reviews on all of your PRs, visit the Cursor dashboard to activate Pro and start your 14-day free trial.