uukuguy/speechless-orca-platypus-coig-lite-4k-0.5e-13b

TEXT GENERATIONConcurrency Cost:1Model Size:13BQuant:FP8Ctx Length:4kPublished:Aug 30, 2023License:cc-by-nc-4.0Architecture:Transformer Open Weights Cold

The uukuguy/speechless-orca-platypus-coig-lite-4k-0.5e-13b is a 13 billion parameter Llama 2-based causal language model, fine-tuned with a 4096-token context window. It is a merge of OpenOrca-Platypus2-13B, enhanced with COIG-PC-LITE and OpenOrca datasets to introduce Chinese language capabilities. This model is designed to combine the reasoning strengths of Platypus with the instruction-following of OpenOrca, now with added support for Chinese language tasks.

Loading preview...

Model Overview

uukuguy/speechless-orca-platypus-coig-lite-4k-0.5e-13b is a 13 billion parameter language model built upon the Llama 2 transformer architecture. It is a specialized fine-tune of the OpenOrca/OpenOrca-Platypus2-13B model, which itself is a merge of garage-bAInd/Platypus2-13B and Open-Orca/OpenOrcaxOpenChat-Preview2-13B. The key differentiator for this specific model is its enhanced Chinese language capability, achieved by incorporating 10% COIG-PC-LITE and 10% OpenOrca datasets, alongside 100% Open-Platypus data.

Key Capabilities & Features

  • Multilingual Support: Introduces Chinese language capabilities to the OpenOrca-Platypus lineage.
  • Merged Architecture: Combines the strengths of Platypus2-13B (known for STEM and logic tasks) and OpenOrcaxOpenChat-Preview2-13B (instruction-tuned on GPT-4 data).
  • Context Window: Supports a context length of 4096 tokens.
  • Performance: The base OpenOrca-Platypus2-13B model demonstrated strong performance on benchmarks, achieving an average of 64.56 on the HuggingFace Leaderboard (MMLU 59.5, ARC 62.88, HellaSwag 83.19, TruthfulQA 52.69). It also showed 112% of the base model's performance on AGIEval and 105% on BigBench-Hard.

Ideal Use Cases

This model is particularly well-suited for applications requiring a 13B parameter model with strong instruction-following and reasoning abilities, especially when Chinese language understanding and generation are critical. It can be leveraged for tasks that benefit from a blend of logical reasoning and comprehensive instruction-tuning.

Finetunes
2 models