You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Descriptions:
Currently, we only support basic interventions during model generation just like the model forward call. This is not ideal. In model generation, we want to support more free-formed interventions (e.g., intervene based on decoding steps or other decoding parameters, not just unit location as if it is in an intervened forward mode).
The current infra (also this applies to other existing intervention library as well) cannot support this. For instance, it does not support a specific decoding step intervention during decoding without more incisive code change. To support complex cases, we plan to introduce a new notion of Intervention Scheduler.
In the high-level, the scheduler is responsible to schedule interventions dynamically at inference time, and it is customizable. For instance, we can (1) intervene on all decoded punctuation tokens, or (2) all verbs that get decoded, or (3) all the last entity token that gets decoded in a specific entity set.
This enables us to a wide spectrum of ways to steer model behavior with interventions. This ticket may require multiple changes.
The text was updated successfully, but these errors were encountered:
Descriptions:
Currently, we only support basic interventions during model generation just like the model forward call. This is not ideal. In model generation, we want to support more free-formed interventions (e.g., intervene based on decoding steps or other decoding parameters, not just unit location as if it is in an intervened forward mode).
The current infra (also this applies to other existing intervention library as well) cannot support this. For instance, it does not support a specific decoding step intervention during decoding without more incisive code change. To support complex cases, we plan to introduce a new notion of Intervention Scheduler.
In the high-level, the scheduler is responsible to schedule interventions dynamically at inference time, and it is customizable. For instance, we can (1) intervene on all decoded punctuation tokens, or (2) all verbs that get decoded, or (3) all the last entity token that gets decoded in a specific entity set.
This enables us to a wide spectrum of ways to steer model behavior with interventions. This ticket may require multiple changes.
The text was updated successfully, but these errors were encountered: