# Third-Party Model and Code Notices This file records upstream identifiers and the licence metadata observed during package assembly. It is an inventory, not legal advice and not a substitute for the recipient's licence review. ## Qwen2.5-Coder Models - Upstream organisation: Qwen / Alibaba Cloud. - Relevant upstreams: `Qwen/Qwen2.5-Coder-7B`, `Qwen/Qwen2.5-Coder-0.5B`, and `Qwen/Qwen2.5-Coder-0.5B-Instruct`. - The official 7B model metadata declares `apache-2.0`. - The upstream Apache License 2.0 text captured with revision `0396a76181e127dfc13e5c5ec48a8cee09938b02` is included as `licenses/QWEN2.5-CODER-7B-APACHE-2.0.txt`. - Official source: ## GPT-2 - Upstream: `openai-community/gpt2`. - Fixed upstream revision: `607a30d783dfa663caf39e06633721c8d4cfcd7e`. - The official Hugging Face model metadata declares `license: mit`. - Official source: ## CodeGPT-Multilingual — Rights Review Required - Upstream: `AISE-TUDelft/CodeGPT-Multilingual`. - Fixed upstream revision: `f373547d1234992a64cd08d64cd69abbb3a83579`. - At package assembly, the official model metadata had no licence tag and the repository file list contained no `LICENSE` file. - The delivered continued-pre-training checkpoint is therefore **not marked as cleared for external redistribution**. Obtain written rights confirmation or a licence declaration before transferring it beyond the authorised delivery channel. - Official source: ## llm.c Training Code The included llm.c training-code subset is under the licence text copied to `code/llmc_training/LICENSE`. The preserved source snapshot corresponds to Git commit `f1e2ace651495b74ae22d45d1723443fd00ecd3a`.