Showing 2 open source projects for "decoder"

View related business solutions
  • Data management solutions for confident marketing Icon
    Data management solutions for confident marketing

    For companies wanting a complete Data Management solution that is native to Salesforce

    Verify, deduplicate, manipulate, and assign records automatically to keep your CRM data accurate, complete, and ready for business.
    Learn More
  • Failed Payment Recovery for Subscription Businesses Icon
    Failed Payment Recovery for Subscription Businesses

    For subscription companies searching for a failed payment recovery solution to grow revenue, and retain customers.

    FlexPay’s innovative platform uses multiple technologies to achieve the highest number of retained customers, resulting in reduced involuntary churn, longer life span after recovery, and higher revenue. Leading brands like LegalZoom, Hooked on Phonics, and ClinicSense trust FlexPay to recover failed payments, reduce churn, and increase customer lifetime value.
    Learn More
  • 1
    Basaran

    Basaran

    Basaran, an open-source alternative to the OpenAI text completion API

    ...The open source community will eventually witness the Stable Diffusion moment for large language models (LLMs), and Basaran allows you to replace OpenAI's service with the latest open-source model to power your application without modifying a single line of code. Stream generation using various decoding strategies. Support both decoder-only and encoder-decoder models. Detokenizer that handles surrogates and whitespace. Multi-GPU support with optional 8-bit quantization. Real-time partial progress using server-sent events. Compatible with OpenAI API and client libraries. Comes with a fancy web-based playground. Docker images are available on Docker Hub and GitHub Packages.
    Downloads: 0 This Week
    Last Update:
    See Project
  • 2
    CPT

    CPT

    CPT: A Pre-Trained Unbalanced Transformer

    A Pre-Trained Unbalanced Transformer for Both Chinese Language Understanding and Generation. We replace the old BERT vocabulary with a larger one of size 51271 built from the training data, in which we 1) add missing 6800+ Chinese characters (most of them are traditional Chinese characters); 2) remove redundant tokens (e.g. Chinese character tokens with ## prefix); 3) add some English tokens to reduce OOV. Position Embeddings We extend the max_position_embeddings from 512 to 1024. We...
    Downloads: 0 This Week
    Last Update:
    See Project
  • Previous
  • You're on page 1
  • Next