wesley7137/Neuro-Sci-PyWizCoder-13B-V1-merged

TEXT GENERATIONConcurrency Cost:1Model Size:13BQuant:FP8Ctx Length:4kArchitecture:Transformer Cold

The wesley7137/Neuro-Sci-PyWizCoder-13B-V1-merged model is a 13 billion parameter language model with a 4096 token context length. This model was trained using bitsandbytes 4-bit quantization, specifically nf4, and PEFT 0.4.0. Its primary differentiator and use case are not explicitly detailed in the provided README, which focuses on training configuration.

Loading preview...

Model Overview

This model, wesley7137/Neuro-Sci-PyWizCoder-13B-V1-merged, is a 13 billion parameter language model. The provided information primarily details its training configuration rather than specific capabilities or intended use cases.

Training Details

The model was trained utilizing bitsandbytes for quantization, specifically employing a 4-bit quantization scheme (nf4). Key parameters for this quantization included:

  • load_in_4bit: True
  • bnb_4bit_quant_type: nf4
  • bnb_4bit_compute_dtype: float16

Additionally, the training process incorporated the PEFT (Parameter-Efficient Fine-Tuning) framework, with version 0.4.0 being used.

Key Characteristics

  • Parameter Count: 13 billion parameters.
  • Context Length: 4096 tokens.
  • Quantization: Trained with bitsandbytes 4-bit nf4 quantization.
  • Framework: Utilizes PEFT 0.4.0 for efficient fine-tuning.

Use Cases

Based solely on the provided README, specific optimized use cases are not detailed. Users should infer potential applications based on the model's size and general language model capabilities, considering the training methodology focused on efficient quantization.