madatnlp/mist-enko-lora-2950

TEXT GENERATIONConcurrency Cost:1Model Size:7BQuant:FP8Ctx Length:8kPublished:Dec 16, 2023License:apache-2.0Architecture:Transformer Open Weights Cold

The madatnlp/mist-enko-lora-2950 is a 7 billion parameter language model based on the Mistral-7b-v0.1 architecture, further pretrained using a LoRA adapter with a rank of 128 and alpha of 16. This model specializes in translation tasks, having been trained on various translation datasets sourced from ai-hub. It offers a context length of 8192 tokens, making it suitable for applications requiring robust translation capabilities.

Loading preview...

Model Overview

The madatnlp/mist-enko-lora-2950 is a 7 billion parameter language model built upon the Mistral-7b-v0.1 base architecture. It has undergone further pretraining using a LoRA (Low-Rank Adaptation) adapter, configured with a rank of 128 and an alpha value of 16.

Key Capabilities

  • Translation Focus: The model's primary specialization is in translation, having been extensively pretrained on diverse translation datasets from ai-hub.
  • Mistral-7b Foundation: Leverages the strong base capabilities of the Mistral-7b-v0.1 model.
  • Efficient Adaptation: Utilizes LoRA for efficient fine-tuning, allowing for specialized performance without modifying the entire base model.

Good For

  • Translation Tasks: Ideal for applications requiring robust and accurate translation between languages, particularly those covered in the ai-hub datasets used for its pretraining.
  • Resource-Efficient Deployment: As a LoRA-adapted model, it offers a more efficient deployment footprint compared to full model fine-tunes, making it suitable for environments with computational constraints.
  • Research and Development: Provides a strong foundation for further experimentation and fine-tuning on specific translation pairs or domains.

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