Commit Graph

  • 278d87286a Remove broken in-place replacement code, use _tokenize_with_special_tokens only master Jinx 2026-04-10 17:35:09 +00:00
  • ca50973065 Fix tokenization: replace text token sequences with actual special token IDs Jinx 2026-04-10 17:28:51 +00:00
  • d3b5f04f88 Add embed_tokens to LoRA targets + token ID verification before training Jinx 2026-04-10 17:20:06 +00:00
  • f46995690c Update runbook for tool-call token training run Jinx 2026-04-10 17:14:57 +00:00
  • af497eb16c Add training plan: teach model to emit native tool-call tokens Jinx 2026-04-10 17:07:28 +00:00
  • d1e8c306e3 Add critical training objective: teach model to emit native tool-call tokens Jinx 2026-04-10 16:52:09 +00:00
  • eb0850bca6 Fix: correct source dataset tag names for tool_call/tool_response Jinx 2026-04-10 06:37:43 +00:00
  • 99481ca127 Add HF_TOKEN env var Jinx 2026-04-10 06:33:09 +00:00
  • 9e723d393e Fix: remove unsupported --model arg from prepare_data.py call Jinx 2026-04-10 06:32:55 +00:00
  • adbd85366b init commit Jinx 2026-04-10 06:24:05 +00:00
  • 46a3ddbb25 Add deployment runbook Jinx 2026-04-10 05:28:30 +00:00
  • 6af62c85d5 Add docker-compose with /srv mounts for persistent data Jinx 2026-04-10 05:21:04 +00:00
  • 82348341b0 Initial LoRA training setup for SmolLM3-3B tool calling Jinx 2026-04-10 05:11:05 +00:00