[manus] Modular DPO Dataset Builder for MLX (#5) #6
Closed
manus
wants to merge 1 commits from
manus/dpo-data-pipeline into main
pull from: manus/dpo-data-pipeline
merge into: Timmy_Foundation:main
Timmy_Foundation:main
Timmy_Foundation:gemini/issue-182
Timmy_Foundation:master
Timmy_Foundation:feat/architecture-linter-provenance
Timmy_Foundation:feat/adr-system-provenance
Timmy_Foundation:sonnet/smoke-test-sonnet
Timmy_Foundation:sonnet/issue-260
Timmy_Foundation:docs/automation-audit-20260404
Timmy_Foundation:docs/architecture-kt-unified-schema
Timmy_Foundation:feat/frontier-local-layer-4-mesh
Timmy_Foundation:timmy/code-claw-docs
Timmy_Foundation:claw-code/issue-232
Timmy_Foundation:feat/frontier-local-layer-5-immortality
Timmy_Foundation:feat/frontier-local-layer-3
Timmy_Foundation:feature/workforce-manager
Timmy_Foundation:feat/frontier-local-agenda-v2
Timmy_Foundation:feat/cost-saving-guide
Timmy_Foundation:timmy/gemini-loop-hardening
Timmy_Foundation:timmy/orchestrator-kimi-heartbeat-status
Timmy_Foundation:timmy/orchestrator-kimi-visibility
Timmy_Foundation:timmy/issue-186-import-bridge
Timmy_Foundation:codex/workflow-pr-review
Timmy_Foundation:feat/sovereign-identity-phase-23
Timmy_Foundation:feat/sovereign-evolution-redistribution
Timmy_Foundation:gemini/orchestration-hardening
Timmy_Foundation:gemini/audit-bugfixes
Timmy_Foundation:timmy/issue-86-z3-crucible
Timmy_Foundation:feat/allegro-identity-fix
Timmy_Foundation:gemini/issue-75
Timmy_Foundation:gemini/issue-76
Timmy_Foundation:gemini/issue-78
Timmy_Foundation:review/move-last-two-main-commits-20260328-000322
Timmy_Foundation:gemini/issue-50
Timmy_Foundation:backup/main-before-reset-20260328-000322
Timmy_Foundation:gemini/issue-52
Timmy_Foundation:gemini/issue-54
Timmy_Foundation:fix/mcp-morrowind-tool-naming
Timmy_Foundation:gemini/issue-59
Timmy_Foundation:gemini/issue-60
Timmy_Foundation:gemini/issue-61
Timmy_Foundation:gemini/issue-62
Timmy_Foundation:gemini/issue-63
Timmy_Foundation:gemini/issue-41
Timmy_Foundation:gemini/issue-42
Timmy_Foundation:gemini/issue-43
Timmy_Foundation:codex/hermes-venv-runner
Timmy_Foundation:codex/twitter-archive-orchestration
Timmy_Foundation:codex/cleanup-pass-2
Timmy_Foundation:codex/cleanup-boundaries
Timmy_Foundation:gemini/issue-8
Timmy_Foundation:gemini/issue-20
Timmy_Foundation:gemini/issue-21
Timmy_Foundation:gemini/issue-22
Timmy_Foundation:gemini/issue-9
Timmy_Foundation:gemini/issue-10
Timmy_Foundation:gemini/issue-11
Timmy_Foundation:gemini/issue-12
Timmy_Foundation:gemini/issue-13
Timmy_Foundation:feature/dpo-training-pipeline
1 Commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
|
|
2fe6e33c05 |
feat: implement modular DPO dataset builder for MLX (#5)
- Created training/build_dpo_pairs.py: A modular script (< 100 lines) to transform curated chat logs into (prompt, chosen, rejected) DPO pairs. - Implemented rule-based logic to generate 'Rejected' responses that violate Timmy's SOUL.md values (verbosity, corporate tone, disclaimers). - Verified the output schema against mlx-lm requirements. - Generated a local DPO_REPORT.md with validation metrics. - unblocks Issue #5: DPO training on MLX. |