CobraMamba/mamba-gpt-7b-v1

TEXT GENERATIONConcurrency Cost:1Model Size:7BQuant:FP8Ctx Length:8kPublished:Oct 13, 2023License:apache-2.0Architecture:Transformer0.0K Open Weights Cold

CobraMamba/mamba-gpt-7b-v1 is a 7 billion parameter instruction-tuned causal language model based on the OpenLLaMA architecture, developed by CobraMamba. Fine-tuned on diverse datasets including Stanford Alpaca and UltraChat, it demonstrates performance comparable to larger 7B models. This model is optimized for general-purpose conversational AI and instruction following, making it suitable for a wide range of natural language processing tasks.

Loading preview...

CobraMamba/mamba-gpt-7b-v1: An Instruction-Tuned OpenLLaMA Model

CobraMamba/mamba-gpt-7b-v1 is a 7 billion parameter language model built upon the openlm-research/open_llama_7b_v2 base model. This model has undergone extensive instruction fine-tuning across a variety of datasets, including Stanford Alpaca, Open Assistant (multilingual), LIMA, CodeAlpaca 20k, GPT-4 Generated Data, and UltraChat.

Key Capabilities & Performance

  • Enhanced Performance: Through fine-tuning, mamba-gpt-7b-v1 surpasses the original OpenLLaMA model in various evaluation subtasks, achieving performance comparable to other 7B models.
  • Instruction Following: Optimized for understanding and executing instructions from diverse sources.
  • Multilingual Data Exposure: Training on datasets like Open Assistant (multilingual) contributes to its versatility.
  • Benchmark Results: On the Open LLM Leaderboard, it achieves an average score of 54.77, with notable scores including 61.26 on ARC (25-shot) and 84.1 on HellaSwag (10-shot).

Good For

  • General-purpose conversational AI applications.
  • Instruction-based text generation and completion.
  • Tasks requiring understanding and response generation based on diverse prompts.
  • Developers seeking a capable 7B model with strong instruction-following abilities.

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