MarinaraSpaghetti/Nemomix-v1.0-12B

TEXT GENERATIONPricing:Input $0.87 / Cached $0.2 / Output $0.99Concurrent Unit Cost:1Model Size:12BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Jul 28, 2024Architecture:Transformer0.0K Featherless Exclusive Cold

MarinaraSpaghetti/Nemomix-v1.0-12B is a 12 billion parameter language model merged from several base models, including Mistral-Nemo-Instruct-2407, Lumimaid-v0.2-12B, mini-magnum-12b-v1.1, and shuttle-2.5-mini. This model aims to combine the instructional intelligence of base Nemo models with enhanced prose quality for roleplaying and creative text generation. It supports both Mistral Instruct and ChatML formats and is optimized for nuanced text generation with specific temperature and Min P settings.

Loading preview...

Nemomix-v1.0-12B Overview

Nemomix-v1.0-12B is a 12 billion parameter language model developed by MarinaraSpaghetti, created through a merge of several pre-trained models using the DARE TIES merge method. The primary goal of this version is to integrate the instructional capabilities of the base Nemo model with the improved prose and creative writing styles found in various roleplaying fine-tunes.

Key Capabilities

  • Blended Intelligence and Prose: Combines the "smartness" of base Instruct Nemo with the "better prose" from roleplaying fine-tunes.
  • Merge Architecture: Built upon a base of Mistral-Nemo-Base-2407, incorporating contributions from:
    • intervitens_mini-magnum-12b-v1.1
    • mistralaiMistral-Nemo-Instruct-2407
    • NeverSleep_Lumimaid-v0.2-12B
    • shuttleai_shuttle-2.5-mini
  • Instruction Format Compatibility: Supports both Mistral Instruct and ChatML formats for flexible prompting.
  • Optimized Generation Settings: Recommended settings include a lower temperature of 0.35, with potential for higher temperatures (1.0-1.2) when combined with increased Min P (0.01-0.1) and specific DRY parameters (0.8/1.75/2/0).

Good For

  • Use cases requiring a balance between intelligent instruction following and high-quality, creative prose.
  • Applications involving roleplaying, story generation, or other forms of nuanced text creation.
  • Developers familiar with Mistral Instruct or ChatML prompting styles.

Finetunes
3 models