1.3b / README.md
Corianas's picture
Adding Evaluation Results (#2)
07ec0fe
metadata
license: cc-by-nc-4.0
datasets:
  - tatsu-lab/alpaca
language:
  - en

Model Card for Cerebras 1.3b Dollyfied.

This is a finetuned model of Cerebras 1.3B model using DataBricksLabs Dolly Framework

Example output:

Instruction: Write a love letter to Edgar Allan Poe.

Edgar Allan Poe is here again, with a brand new treasure of tales for you to enjoy. First, a man's life as a fallen angel, then his lonely falls into the sea, and finally a journey through the wild, exciting lands of the American West. He also carries a mysterious message for us: never stop searching for our own sweet. Thank you, Edgar Allan Poe!


Instruction: Write a tweet announcing Dolly, a large language model from Databricks.

Dolly a brand new model from Databricks that has been thoroughly validated and improved with feedback from experts and users! #databricks #languagemodels #databrickslanguage


Instruction: I'm selling my Nikon D-750, write a short blurb for my ad.

We bring you the revolutionary Nikon D-750 - the perfect camera for any shoot. With the Nikon D-750, you'll be able to achieve stunning images with state-of-the-art features and brilliant images. The Nikon D-750 is the perfect camera for everyday work, and it produces brilliant, vibrant images with spectacular clarity. Whether you’re a professional, a hobbyist, or an enthusiast, the Nikon D-750 is the perfect camera for you. Get yours today!


Instruction: Explain to me the difference between nuclear fission and fusion.

Nuclear fission and fusion are both processes that use nuclear material and energy to generate heat. Fission, which occurs at high enough temperatures, produces either a proton or a neutron with the same energy level. Fusion, on the other hand, relies on a hydrogen atom and is considered to be the most stable process. Nuclear fission and fusion produce different amounts of heat, which is why nuclear fusion can be used to generate heat energy for a variety of applications.


Instruction: Give me a list of 5 science fiction books I should read next.

  1. The coming of the Red Planet by Guy Adams
  2. The Hitchhiker's Guide to the Galaxy by Douglas Adams
  3. Out of the Ashes by Douglas Adams
  4. The Hitchhiker's Guide to the Galaxy episode "The Ancient of Days"
  5. Icy Girls by J.R.R. Tolkien

Model Details

Model Description

This is a finetuned version of cerebras' 1.3Billion paramater model that has been trained to follow instructions.

It was accomplished using DataBricks Dolly training tools.

Uses

This is a simple GPT chatbot that has been finetuned to understand instructions. Its knowledge about facts about the world is should be considered suspect at best.

Direct Use

[More Information Needed]

Downstream Use [optional]

[More Information Needed]

Out-of-Scope Use

[More Information Needed]

Bias, Risks, and Limitations

[More Information Needed]

Recommendations

Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.

How to Get Started with the Model

Use the code below to get started with the model.

[More Information Needed]

Training Details

Training Data

[More Information Needed]

Training Procedure

Preprocessing [optional]

[More Information Needed]

Training Hyperparameters

  • Training regime: [More Information Needed]

Speeds, Sizes, Times [optional]

[More Information Needed]

Evaluation

Testing Data, Factors & Metrics

Testing Data

[More Information Needed]

Factors

[More Information Needed]

Metrics

[More Information Needed]

Results

[More Information Needed]

Summary

Model Examination [optional]

[More Information Needed]

Environmental Impact

Carbon emissions can be estimated using the Machine Learning Impact calculator presented in Lacoste et al. (2019).

  • Hardware Type: 8xA100
  • Hours used: 30mins
  • Cloud Provider: Lambda GPU

Technical Specifications [optional]

Model Architecture and Objective

GPT2 Cerebras-GPT 1.3B Layers 24
n_embd 2048 Heads 16

Hardware

8xA100s

Software

https://github.com/databrickslabs/dolly

Model Card Contact

[More Information Needed]

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric Value
Avg. 27.02
ARC (25-shot) 27.3
HellaSwag (10-shot) 38.3
MMLU (5-shot) 26.77
TruthfulQA (0-shot) 39.02
Winogrande (5-shot) 53.04
GSM8K (5-shot) 0.15
DROP (3-shot) 4.57