Changing batch_size after Pipeline initialization?

[asusdisciple]

So I know this is a very basic question, but I am not sure how I can change the batch_size - and parameters in general for pipelines and the models, which are used in the pipeline AFTER it is initialized! Is there a tutorial on this? I know how to load them from a .yaml with:

pipe =  Pipeline.from_pretrained("pyannote/speaker-diarization-3.1",
                                                          use_auth_token="xxxx",
                                                          cache_dir="models",
                                                          hparams_file="models/config.yaml")

However I found it very hard to change parameters after this because pipe.parameters() only gets the current configuration. Also when I instantiate a pipeline I found this code in the tutorial:

finetuned_pipeline.instantiate({
    "segmentation": {
        "threshold": best_segmentation_threshold,
        "min_duration_off": 0.0,
    },
    "clustering": {
        "method": "centroid",
        "min_cluster_size": 15,
        "threshold": best_clustering_threshold,
    },
})

However is there any documentation where I can find a list of usable hyperparameters? Like a big dummy dictionary, which tells me how to set them? In the case above for example I do not know how to set the batch_size. Do I Have to add a new key "Pipeline" with this parameter nested in another dict? I also tried adding batch_size_segmentation as in the config.yaml in the segmentation dict but it did not work.

Any help would be appreciated :)

[hbredin]

pipeline.segmentation_batch_size = 32 should do the trick.

[asusdisciple]

Thanks a lot! This did the trick indeed.