askmyteapot/metharme

TEXT GENERATIONConcurrency Cost:1Model Size:7BQuant:FP8Ctx Length:4kArchitecture:Transformer Cold

askmyteapot/metharme is a 7 billion parameter language model based on the Metharme architecture, utilizing TehVenom/Metharme-7b-Merged-Safetensors as its base. This model is quantized and provides specific perplexity scores on C4, Wiki2, and PTB datasets, indicating its performance characteristics. It is suitable for tasks requiring a compact yet capable language model with a 4096-token context length.

Loading preview...

Metharme 7B Overview

askmyteapot/metharme is a 7 billion parameter language model built upon the TehVenom/Metharme-7b-Merged-Safetensors base. This version has undergone quantization, making it more efficient for deployment and inference, particularly in resource-constrained environments.

Key Characteristics

  • Base Model: Utilizes the TehVenom/Metharme-7b-Merged-Safetensors as its foundational architecture.
  • Quantization: The model has been quantized, which typically reduces its memory footprint and speeds up inference.
  • Performance Metrics: Perplexity scores are provided for standard benchmarks:
    • C4: 7.8935709 (Act-order and True-Sequential) and 8.104910851 (True-Sequential Groupsize 32)
    • Wiki2: 6.309680939 (Act-order and True-Sequential) and 7.025843143 (True-Sequential Groupsize 32)
    • PTB: 46.60740662 (Act-order and True-Sequential) and 48.19175339 (True-Sequential Groupsize 32)
  • Context Length: Supports a context window of 4096 tokens.

Good For

  • Applications requiring a 7B parameter model with reduced memory usage due to quantization.
  • Tasks where the provided perplexity scores align with desired performance benchmarks.
  • Environments where efficient inference is a priority.