Skip to content

[FEATURE]: train seq2seq model  #3564

Closed

Description

@Sahajtomar

Describe the feature

Is it not possible yet to train seq2seq model using colossalai? Given scripts are for causalLM models

Activity

  1. Issues-translate-bot commented on Apr 14, 2023

    @Issues-translate-bot

    Bot detected the issue body's language is not English, translate it automatically. 👯👭🏻🧑‍🤝‍🧑👫🧑🏿‍🤝‍🧑🏻👩🏾‍🤝‍👨🏿👬🏿


    Title: [FEATURE]: train seq2seq model

  2. binmakeswell commented on Apr 18, 2023

    @binmakeswell
    Member

    @Sahajtomar This is just an example, welcome to try other models and submit PRs. For example #3567 Thanks.

  3. JazyJiang commented on Oct 19, 2026

    @JazyJiang
    Collaborator

    To follow up: seq2seq models can be trained with ColossalAI. The Booster plugins work with any torch.nn.Module, so an encoder-decoder model (e.g. T5) is boosted the same way as a causal LM; the provided scripts are only examples. Closing as answered.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    enhancementNew feature or request

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions