RichardErkhov/InferenceIllusionist_-_Excalibur-7b-DPO-gguf

Quantization made by Richard Erkhov.

Github

Discord

Request more models

Excalibur-7b-DPO - GGUF

• —Model creator: https://huggingface.co/InferenceIllusionist/
• —Original model: https://huggingface.co/InferenceIllusionist/Excalibur-7b-DPO/

Original model description: --- license: apache-2.0 library_name: transformers tags:

• —finetune
• —dpo
• —chatml base_model:
• —InferenceIllusionist/Excalibur-7b datasets:
• —Intel/orcadpopairs model-index:
• —name: Excalibur-7b-DPO results:
• —task: type: text-generation name: Text Generation dataset: name: AI2 Reasoning Challenge (25-Shot) type: ai2arc config: ARC-Challenge split: test args: numfew_shot: 25 metrics:
• —type: accnorm value: 70.9 name: normalized accuracy source: url: https://huggingface.co/spaces/HuggingFaceH4/openllm_leaderboard?query=InferenceIllusionist/Excalibur-7b-DPO name: Open LLM Leaderboard
• —task: type: text-generation name: Text Generation dataset: name: HellaSwag (10-Shot) type: hellaswag split: validation args: numfewshot: 10 metrics:
• —type: accnorm value: 87.93 name: normalized accuracy source: url: https://huggingface.co/spaces/HuggingFaceH4/openllm_leaderboard?query=InferenceIllusionist/Excalibur-7b-DPO name: Open LLM Leaderboard
• —task: type: text-generation name: Text Generation dataset: name: MMLU (5-Shot) type: cais/mmlu config: all split: test args: numfewshot: 5 metrics:
• —type: acc value: 65.46 name: accuracy source: url: https://huggingface.co/spaces/HuggingFaceH4/openllmleaderboard?query=InferenceIllusionist/Excalibur-7b-DPO name: Open LLM Leaderboard
• —task: type: text-generation name: Text Generation dataset: name: TruthfulQA (0-shot) type: truthfulqa config: multiplechoice split: validation args: numfewshot: 0 metrics:
• —type: mc2 value: 70.82 source: url: https://huggingface.co/spaces/HuggingFaceH4/openllmleaderboard?query=InferenceIllusionist/Excalibur-7b-DPO name: Open LLM Leaderboard
• —task: type: text-generation name: Text Generation dataset: name: Winogrande (5-shot) type: winogrande config: winograndexl split: validation args: numfew_shot: 5 metrics:
• —type: acc value: 82.48 name: accuracy source: url: https://huggingface.co/spaces/HuggingFaceH4/openllmleaderboard?query=InferenceIllusionist/Excalibur-7b-DPO name: Open LLM Leaderboard
• —task: type: text-generation name: Text Generation dataset: name: GSM8k (5-shot) type: gsm8k config: main split: test args: numfewshot: 5 metrics:
• —type: acc value: 65.43 name: accuracy source: url: https://huggingface.co/spaces/HuggingFaceH4/openllmleaderboard?query=InferenceIllusionist/Excalibur-7b-DPO name: Open LLM Leaderboard ---

Excalibur-7b-DPO

<img src="https://i.imgur.com/pbPbqq0.jpeg" width="550"/>

An initial foray into the world of fine-tuning. The goal of this release was to amplify the quality of the original model's responses, in particular for vision use cases*

<b>Weighted (Importance Matrix) Quants available here</b>

<b>Static (Legacy) quants available here</b>

Notes & Methodology

• —Excalibur-7b fine-tuned with Direct Preference Optimization (DPO) using Intel/orcadpopairs
• —This is a quick experiment to determine the impact of DPO finetuning on the Excelsior-7b base model
• —Ran for a little over an hour on a single A100
• —Fine-tuning succeeded in making model conversational and more well-rounded
• —Benchmark scores increased in the following categories versus base Excelsior-7b:
• —ARC: 69.71 -> <b>70.9</b>
• —HellaSwag: 87.56 -> <b>87.93</b>
• —TruthfulQA: 67.24 -> <b>70.82</b>
• —Average: 73.6 -> <b>73.84</b>
• —Precision: bfloat16

Sample Question - Vision

<img src="https://i.imgur.com/7aRWtzU.jpeg" width="425"/>

*<b>Requires additional mmproj file. You have two options for vision functionality (available inside this repo):</b>

• —Quantized - Limited VRAM Option (197mb)
• —Unquantized - Premium Option / Best Quality (596mb)

Select the gguf file of your choice in Koboldcpp as usual, then make sure to choose the mmproj file above in the LLaVA mmproj field of the model submenu: <img src="https://i.imgur.com/x8vqH29.png" width="425"/>

Prompt Format

• —For best results please use ChatML for the prompt format. Alpaca may also work.

Open LLM Leaderboard Evaluation Results

Detailed results can be found here