|
Download README.md from CompiwerAI/Mtrini-Decision-1.0: direct link, hf CLI and curl.
- Browser
- Download file 1.98 kB
-
https://huggingface.co/CompiwerAI/Mtrini-Decision-1.0/resolve/main/README.md
- Command line
-
hf download hf://CompiwerAI/Mtrini-Decision-1.0/README.md
-
curl -L -o README.md https://huggingface.co/CompiwerAI/Mtrini-Decision-1.0/resolve/main/README.md
1.98 kB
| license: other | |
| base_model: Qwen/Qwen3-8B | |
| tags: | |
| - compiwer-ai | |
| - mtrini | |
| - decision-making | |
| - research | |
| - experimental | |
| - lora | |
| - qwen3 | |
| # 🧠 Mtrini-Decision-1.0 | |
| **An experimental decision-model research project by Compiwer AI.** | |
| Mtrini-Decision explores specialized decision-making with a Qwen3-8B | |
| backbone, a LoRA adapter, and a learned decision head. | |
| > **Status: Experimental research preview.** | |
| ## Model details | |
| - **Developer:** Compiwer AI | |
| - **Base model:** [Qwen/Qwen3-8B](https://huggingface.co/Qwen/Qwen3-8B) | |
| - **Checkpoint:** Training step 100 | |
| - **Components:** LoRA adapter and `decision_head.pt` | |
| - **Research goals:** structured choices, option ranking, and decision scoring | |
| ## Files | |
| - `adapter_config.json`: LoRA configuration | |
| - `adapter_model.safetensors`: LoRA adapter weights | |
| - `decision_head.pt`: custom decision-head checkpoint | |
| - `mtrini_decision_config.json`: model configuration | |
| - `release_metrics.json`: release status and caveats | |
| ## Inference | |
| This repository contains a custom decision head. Loading the LoRA adapter | |
| with a standard Transformers/PEFT workflow does not automatically run the | |
| custom head. | |
| To reproduce decision outputs, users need the exact decision-head class, | |
| forward-pass logic, and preprocessing from the training implementation. | |
| Those components must match the saved checkpoint. | |
| ## Evaluation and limitations | |
| This is an early research checkpoint, not a claim of state-of-the-art | |
| performance or production readiness. Initial evaluation metrics require | |
| verification, including target semantics, normalization, and baseline | |
| construction. No validated benchmark result is claimed here. | |
| Do not use this experimental model as the sole basis for high-stakes decisions. | |
| ## License | |
| The license is marked `other` pending verification of redistribution rights | |
| for the base model, training data, and included artifacts. Review those | |
| licenses before redistribution or commercial use. | |
| ## About Compiwer AI | |
| **Building AI For Everyone.** | |