Skip to content

Commit

Permalink
Merge pull request #369 from h2oai/plano-Aug-6-2024
Browse files Browse the repository at this point in the history
LLM Studio tutorial changes. Backbone model and Sampling to speed up training
us8945 authored Aug 2, 2024
2 parents 8f4db9d + 87509f2 commit a22660c
Showing 1 changed file with 9 additions and 7 deletions.
Original file line number Diff line number Diff line change
@@ -56,7 +56,7 @@ The typical LLM Studio workflow is to:
### Import the Data

We will import the influencer data directly from a public bucket on AWS S3.

Start by selecting **Import Datasets**

![llm-studio-import-dataset](assets/llm-studio-import-dataset.jpg)
@@ -125,14 +125,16 @@ Once imported, the data set will appear in the **View datasets** tab.

On the experiment page, various experiment settings are available. For now, we will simply adjust the LLM backbone.

For this lab, we will use the smallest available model, ```facebook/opt-125m```, as this will enable the experiment to finish the quickest.
For this lab, we will use the smallest available model, ```h2oai/h2o-danube2-1.8b-chat```, as this will enable the experiment to finish the quickest.

Under the **Experiments** Section:

1. Go to **LLM Backbone**
2. Click on the dropdown
3. Select ```facebook/opt-125m```
4. Click on **Run experiment**
1. Click on the dropdown
2. **Select ```h2oai/h2o-danube2-1.8b-chat ```**

2. Got **Data Sample** and select `0.2`
3. Click on **Run experiment**

![llm-studio-create-experiment](assets/llm-studio-create-experiment.jpg)

@@ -147,7 +149,7 @@ Note: This will take less than 5 minutes. You can click on **Refresh** to get a
Once the experiment is complete, the status will turn green and change to “finished”.

## Task 3: Explore Finished Experiment

Now, we will explore the results of the finished experiment.

**Under View experiments**:
@@ -216,6 +218,6 @@ Instructions to get your Hugging Face API Token
2. [Get a User Access or API token in your Hugging Face profile settings.](https://huggingface.co/settings/tokens?new_token=true)
3. Please create an Access Token with the Role of “write” for use in the lab. You should see a token ```hf_xxxxx``` (old tokens are ```api_XXXXXhttps://huggingface.co/joinXXX``` or ```api_org_XXXXXXX``` ).

If you choose to download the Model, H2O LLM Studio will automatically create model documentation (e.g., model card) that describes how to use this model further.

If you choose to download the Model, H2O LLM Studio will automatically create model documentation (e.g., model card) that describes how to use this model further.

0 comments on commit a22660c

Please sign in to comment.