NalDice/askvox-llama3.3-70b-16bit

Hugging Face
TEXT GENERATIONConcurrency Cost:4Model Size:70BQuant:FP8Ctx Length:32kPublished:Jan 31, 2025License:apache-2.0Architecture:Transformer Open Weights Warm

NalDice/askvox-llama3.3-70b-16bit is a 70 billion parameter Llama 3.3 model developed by NalDice. This model was fine-tuned from unsloth/llama-3.3-70b-instruct-bnb-4bit, leveraging Unsloth and Huggingface's TRL library for accelerated training. It is optimized for instruction-following tasks, benefiting from the efficiency gains of the Unsloth framework.

Loading preview...

Overview

NalDice/askvox-llama3.3-70b-16bit is a 70 billion parameter Llama 3.3 model, developed by NalDice. It is a fine-tuned variant of the unsloth/llama-3.3-70b-instruct-bnb-4bit base model, specifically optimized for instruction-following capabilities.

Key Characteristics

  • Base Model: Fine-tuned from unsloth/llama-3.3-70b-instruct-bnb-4bit, indicating a foundation in the Llama 3.3 architecture.
  • Training Efficiency: The model's training process was significantly accelerated, reportedly 2x faster, by utilizing the Unsloth library in conjunction with Huggingface's TRL (Transformer Reinforcement Learning) library.
  • Parameter Count: Features 70 billion parameters, placing it in the large-scale language model category.

Intended Use

This model is suitable for applications requiring robust instruction-following performance, leveraging the efficiencies gained from its Unsloth-powered fine-tuning. Its large parameter count suggests strong general-purpose language understanding and generation capabilities.

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