databricks-academy / large-language-models Goto Github PK

Notebooks for Large Language Models (LLMs) Specialization

License: Other

Python 100.00%

large-language-models's Introduction

Large Language Models

This repo contains the notebooks and slides for the Large Language Models: Application through Production course on edX & Databricks Academy.

Notebooks

How to Import the Repo into Databricks?

You first need to add Git credentials to Databricks. Refer to documentation here.
Click Repos in the sidebar. Click Add Repo on the top right.
Clone the "HTTPS" URL from GitHub, or copy https://github.com/databricks-academy/large-language-models.git and paste into the box Git repository URL. The rest of the fields, i.e. Git provider and Repository name, will be automatically populated. Click Create Repo on the bottom right.

How to Import the files from `.dbc` releases on GitHub

You can download the notebooks from a release by navigating to the releases section on the GitHub page:
From the releases page, download the .dbc file. This contains all of the course notebooks, with the structure and meta data.
In your Databricks workspace, navigate to the Workspace menu, click on Home and select Import:
Using the import tool, navigate to the location on your computer where the .dbc file was dowloaded from Step 1. Once you select the file, click Import, and the files will be loaded and extracted to your workspace:

Cluster settings

Which Databricks cluster should I use?

First, select Single Node
This courseware has been tested on Databricks Runtime 13.3 LTS for Machine Learning. If you do not have access to a 13.3 LTS ML Runtime cluster, you will need to install many additional libraries (as the ML Runtime pre-installs many commonly used machine learning packages), and this courseware is not guaranteed to run.

For all of the notebooks except LLM 04a - Fine-tuning LLMs and LLM04L - Fine-tuning LLMs Lab, you can run them on a CPU just fine. We recommend either i3.xlarge or i3.2xlarge (i3.2xlarge will have slightly faster performance).

For these notebooks: LLM 04a - Fine-tuning LLMs and LLM04L - Fine-tuning LLMs Lab, you will need the Databricks Runtime 13.3 LTS for Machine Learning with GPU.

Select GPU instance type of g5.2xlarge.

Install datasets and models

How do I install the datasets and models locally?

To improve performance of the code, we highly recommend pre-installing the datasets and models by running the LLM 00a - Install Datasets notebook.
You should run this script before running any of the other notebooks. This can take up to 25mins to complete.

Slides

Where do I download course slides?

Please click the latest version under the Releases section. You will be able to download the slides in PDF.

large-language-models's People

Contributors

Stargazers

Watchers

Forkers

miguelperalvo balbarka rctruta dsninja kartikmehta09 ikeine beartell yashraj5 samraymond itsshaikaslam oshibuki rosienguyen ki-rin saibaldasprivate zhuohuwu0603 prabhu-kesavan githubaliba aphansal123 aditikhare007 peggyfan veejayendraa dvkuznietsov fenilmaisuria pulkitxchadha pejhoo sdonohoo wsilveira-usp syunar neeth04 harshit158 sondywoo grkhcl juanlamadrid20 harryking120903 troybvo laurynasrekasius dong237 odonneluke alvarocardenasc hannapethani azzahra-cloud arpitsardhana seema7an jamilla-az kleeorin calasius girijeshcse caxqueiroz radhakrishnang kalchakra13 amrynsky ramanans wandabwa2004 kmoh19 svaggar ssmaroju jbohning hardwarize caropilardiaz ramiljoaquin nylvam anuraglahon16 pasanthkumar mahimairaja dscott-tantustech avs6 nukhetms porameht j0nd3str0y3r dbogoslovtsev-hse bonsonsm philhetzel mbhushan donkey24 dbuos manujjoshi namapavan abhiagar2019 epps airjoshua sallyhong bichopaulinho volland janevin hamidfard abisheks01 aibibulaatawula mksaraf jhon-murillo ravirao04 ankit-da kumar9676 vishaljayanty haoyuzhang89 ultrasonex lordaouy ankitrajixr daniellekoduru yvainepan salmanmasih

large-language-models's Issues

Module 3 Q and A Lab

This lab uses the CharacterTextSplitter which splits based on a separator not on chunk_size and chunk_overlap we need to either remove those or use the TokenTextSplitter

Missing the "Releases" section for slides and pdf

Hello!

I could not find the "Releases" section in this repo for getting the slides of the course.
Can you please point out where can I find them?
Thanks,
Florin

Deprecated models in LLM 03

In LLM 03, we call models that have now been deprecated by OpanAI so that the commands fail when we run them.

This is the case in the demo where we define a call to text-babbage-001 which OpenAI says should now be replaced with gpt-3.5-turbo-instruct.

Further down in the Agents section, we use

from langchain.llms import OpenAI
llm = OpenAI()

which defaults to calling text-davinci-003 which is also deprecated and should be replaced by gpt-3.5-turbo-instruct.

Error in 'LLM 01L - LLMs with Hugging Face Lab.py'

I get the below error, when i run the code in Jupyter Notebook:

Adding the below line of code solved the problem:

pip install -U accelerate and restart the notebook

Here is the error discussion: https://github.com/huggingface/transformers/issues/23340

How to get the ipynb version?

Hey Team,

It will be great if you can provided the ipynb version.

Thanks,
Subham

Should let the user know about potential issues with map_rerank chain type in 03L

In LLM 03L - Building LLM Chains Lab notebook - Question 3, when the user tries map_rerank chain type, the model we use ('google/flan-t5-large') may not be powerful enough to consistently generate scores that can be parsed by the default parser in the default prompt in order for the results to be ranked.

The student may see error messages like

Could not parse output: [score between 0 and 100]

There are similar reports of this issue on langchain langchain-ai/langchain#3970

Broken dataset in LLM 04

In LLM 04 demo, we call imdb_ds = load_dataset("imdb") as our fine-tuning dataset.
It looks like there was an update to this dataset, and this line will throw an error ExpectedMoreSplits: {'unsupervised'}.

This can be fixed by forcing a re-install of the latest version of Hugging Face's datasets library. However doing so breaks the code further down where it can't find the train and validation splits in the dataset object.

Offline Access

Really comprehensive contents.
Can you please share if we can run all notebooks offline / without Internet connection after initial model download?

And is there any API access needed to run notebooks in this course?

Thank you

DA.reset_lesson() error

When I run

%run ../Includes/Classroom-Setup

for the first time, and I received errors. I looked inside the code, it was at the line "DA.reset_lesson()". I was able to bypass this by commenting out this line, since it was the first time running the code, there's no need to reset anything. But I think it may be helpful to provide the feedback.