LDCC/LDCC-SOLAR-10.7B

TEXT GENERATIONConcurrency Cost:1Model Size:15BQuant:FP8Ctx Length:8kTool Calling:SupportedPublished:Jan 3, 2024License:cc-by-nc-4.0Architecture:Transformer0.0K Open Weights Cold

LDCC-SOLAR-10.7B is a 15 billion parameter language model developed by Lotte Data Communication's AI Technical Team. This model was fine-tuned using the SOLAR learning method, building upon the KoSOLAR-10.7B-v0.1 base model. It is optimized for general language tasks, leveraging its unique training approach to enhance performance.

Loading preview...

Model Overview

LDCC-SOLAR-10.7B is a 15 billion parameter language model developed by the AI Technical Team at Lotte Data Communication. It was fine-tuned using the learning methodology described in the SOLAR paper, building on the yanolja/KoSOLAR-10.7B-v0.1 base model.

Key Characteristics

  • Developer: Lotte Data Communication AI Technical Team.
  • Training Method: Utilizes the SOLAR learning method for fine-tuning.
  • Base Model: Derived from yanolja/KoSOLAR-10.7B-v0.1.
  • Hardware: Trained on A100x4 * 1 hardware.
  • Software: Fine-tuned using a combination of the DeepSpeed library and HuggingFace TRL Trainer / HuggingFace Accelerate.

Usage Notes

When fine-tuning this model, it is recommended to use the tokenizer.json and tokenizer_config.json files from revision v1.1 to ensure compatibility and correct behavior.

Finetunes
3 models