Skip to content

Port Tokenizer and token processor to dfdx #1

@swfsql

Description

@swfsql

To reduce the amount of dependencies (or rather concentrate more on dfdx), the Tokenizer (token encoding/decoding) should be ported to dfdx.

This also applies to the token processor, which makes a token selection based on the Vocab output logits.

Metadata

Metadata

Assignees

No one assigned

    Labels

    enhancementNew feature or request

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions