anakin87/Llama-3-8b-ita-slerp

Hugging Face
TEXT GENERATIONConcurrency Cost:1Model Size:8BQuant:FP8Ctx Length:8kPublished:May 17, 2024License:llama3Architecture:Transformer0.0K Warm

anakin87/Llama-3-8b-ita-slerp is an 8 billion parameter language model, a merge of two Italian Llama 3-based models (LLaMAntino-3-ANITA-8B-Inst-DPO-ITA and Llama-3-8b-Ita) using the SLERP method. This model is specifically designed and optimized for Italian language tasks, achieving an average normalized accuracy of 0.6109 across Italian benchmarks like hellaswag_it, arc_it, and m_mmlu_it. It serves as a specialized option for applications requiring strong performance in Italian language understanding and generation.

Loading preview...

Overview

anakin87/Llama-3-8b-ita-slerp is an 8 billion parameter language model, created by anakin87 through the SLERP merge method using mergekit. This model specifically targets Italian language applications, combining the strengths of two prominent Italian Llama 3-based models: swap-uniba/LLaMAntino-3-ANITA-8B-Inst-DPO-ITA and DeepMount00/Llama-3-8b-Ita.

Key Capabilities

  • Italian Language Proficiency: Optimized for tasks requiring strong understanding and generation in Italian.
  • Merged Architecture: Leverages the SLERP merge method to combine the capabilities of two specialized Italian LLMs.
  • Benchmark Performance: Achieves an average normalized accuracy of 0.6109 on key Italian benchmarks, including:
    • hellaswag_it acc_norm: 0.6879
    • arc_it acc_norm: 0.5714
    • m_mmlu_it 5-shot acc: 0.5732

Good For

  • Italian-centric applications: Ideal for chatbots, content generation, and analysis where Italian is the primary language.
  • Research and Development: Useful for exploring merged model performance and Italian language model capabilities. For detailed comparisons, refer to the Leaderboard for Italian Language Models.

Popular Sampler Settings

Top 3 parameter combinations used by Featherless users for this model. Click a tab to see each config.

temperature
top_p
top_k
frequency_penalty
presence_penalty
repetition_penalty
min_p