I'm working with a multimodal dataset. My input prompt has a text portion and an image. I'm trying to display the image as an image, but can't figure out how. I have a csv file with my image url as a column, my text prompt as a column, and my llm output as a column. I'm trying to follow the instructions here: https://arize.com/docs/ax/prompts/prompt-playground/image-inputs-in-playground#method-1-csv-upload But for step 4 I don't see an option for mapping my columns - how do I do that?
To map your CSV columns (including image URLs) when uploading a dataset in Phoenix Prompt Playground, use the "Create via CSV" option in the Datasets tab. After uploading, you should see a step to map each CSV column to its role (e.g., set your image URL column as the image variable, your text as the prompt, etc.). If you do not see this mapping step, it may be a UI issue or a limitation in your current version鈥攃olumn mapping is expected as per the official documentation (see here).If the mapping UI does not appear, try re-uploading the CSV or updating Phoenix to the latest version. Supported image columns should be mapped as image variables, and you can reference them in your prompt using {{variable_name}}. If the problem persists, it may be a bug or a missing feature in your deployment (source).Would you like more detail or troubleshooting steps?Sources:
- Test image & multimodal inputs | Arize Docs
馃挕聽Hint: Mention RunLLM in the thread for followups.
I'm using the hosted version. I've tried reuploading the csv and I don't see any option to do the mapping step. Can you show an example of the UI for mapping - eg where do I click, what should I see
This is what I see - which of these is the mapping step?
moved to Arize AX Support as this is an AX question. Hopefully someone can get back to you soon!!
