diff --git a/.github/workflows/python-api-superb.yaml b/.github/workflows/python-api-superb.yaml
deleted file mode 100644
index fc409ae0..00000000
--- a/.github/workflows/python-api-superb.yaml
+++ /dev/null
@@ -1,27 +0,0 @@
-# TODO: Merge this with allenNLP to have a single workflow for all docker images.
-name: spacy-docker
-
-on:
-  pull_request:
-    paths:
-      - "docker_images/superb/**"
-jobs:
-  build:
-    runs-on: ubuntu-latest
-    steps:
-      - name: Set up Python ${{ matrix.python-version }}
-        uses: actions/setup-python@v2
-        with:
-          python-version: "3.8"
-      - name: Checkout
-        uses: actions/checkout@v2
-      - name: Set up QEMU
-        uses: docker/setup-qemu-action@v1
-      - name: Set up Docker Buildx
-        uses: docker/setup-buildx-action@v1
-      - name: Install dependencies
-        run: |
-          pip install --upgrade pip
-          pip install pytest pillow httpx
-          pip install -e .
-      - run: RUN_DOCKER_TESTS=1 pytest -sv tests/test_dockers.py::DockerImageTests::test_superb
diff --git a/docker_images/superb/Dockerfile b/docker_images/superb/Dockerfile
deleted file mode 100644
index ec2dc32d..00000000
--- a/docker_images/superb/Dockerfile
+++ /dev/null
@@ -1,37 +0,0 @@
-FROM tiangolo/uvicorn-gunicorn:python3.8
-LABEL maintainer="Omar osanseviero@gmail.com"
-
-# Add any system dependency here
-# RUN apt-get update -y && apt-get install libXXX -y
-RUN apt-get update -y
-RUN apt-get install git -y
-RUN apt-get install libsndfile1 -y
-RUN apt-get install ffmpeg -y
-
-RUN pip install -U pip
-RUN pip install --no-cache-dir torch==1.9
-COPY ./requirements.txt /app
-# Speeding up requirements
-RUN pip install --no-cache-dir -r requirements.txt
-COPY ./prestart.sh /app/
-
-
-# Most DL models are quite large in terms of memory, using workers is a HUGE
-# slowdown because of the fork and GIL with python.
-# Using multiple pods seems like a better default strategy.
-# Feel free to override if it does not make sense for your library.
-ARG max_workers=1
-ENV MAX_WORKERS=$max_workers
-ENV HUGGINGFACE_HUB_CACHE=/data
-ENV TORCH_HOME=/data
-
-# Necessary on GPU environment docker.
-# TIMEOUT env variable is used by nvcr.io/nvidia/pytorch:xx for another purpose
-# rendering TIMEOUT defined by uvicorn impossible to use correctly
-# We're overriding it to be renamed UVICORN_TIMEOUT
-# UVICORN_TIMEOUT is a useful variable for very large models that take more
-# than 30s (the default) to load in memory.
-# If UVICORN_TIMEOUT is too low, uvicorn will simply never loads as it will
-# kill workers all the time before they finish.
-RUN sed -i 's/TIMEOUT/UVICORN_TIMEOUT/g' /gunicorn_conf.py
-COPY ./app /app/app
diff --git a/docker_images/superb/app/__init__.py b/docker_images/superb/app/__init__.py
deleted file mode 100644
index e69de29b..00000000
diff --git a/docker_images/superb/app/main.py b/docker_images/superb/app/main.py
deleted file mode 100644
index c736f699..00000000
--- a/docker_images/superb/app/main.py
+++ /dev/null
@@ -1,96 +0,0 @@
-import functools
-import logging
-import os
-from typing import Dict, Type
-
-from api_inference_community.routes import pipeline_route, status_ok
-from app.pipelines import (
-    AutomaticSpeechRecognitionPipeline,
-    Pipeline,
-    SpeechSegmentationPipeline,
-)
-from starlette.applications import Starlette
-from starlette.middleware import Middleware
-from starlette.middleware.gzip import GZipMiddleware
-from starlette.routing import Route
-
-
-TASK = os.getenv("TASK")
-MODEL_ID = os.getenv("MODEL_ID")
-
-
-logger = logging.getLogger(__name__)
-
-
-# Add the allowed tasks
-# Supported tasks are:
-# - text-generation
-# - text-classification
-# - token-classification
-# - translation
-# - summarization
-# - automatic-speech-recognition
-# - ...
-# For instance
-# from app.pipelines import AutomaticSpeechRecognitionPipeline
-# ALLOWED_TASKS = {"automatic-speech-recognition": AutomaticSpeechRecognitionPipeline}
-# You can check the requirements and expectations of each pipelines in their respective
-# directories. Implement directly within the directories.
-ALLOWED_TASKS: Dict[str, Type[Pipeline]] = {
-    "automatic-speech-recognition": AutomaticSpeechRecognitionPipeline,
-    "speech-segmentation": SpeechSegmentationPipeline,
-}
-
-
-@functools.lru_cache()
-def get_pipeline() -> Pipeline:
-    task = os.environ["TASK"]
-    model_id = os.environ["MODEL_ID"]
-    if task not in ALLOWED_TASKS:
-        raise EnvironmentError(f"{task} is not a valid pipeline for model : {model_id}")
-    return ALLOWED_TASKS[task](model_id)
-
-
-routes = [
-    Route("/{whatever:path}", status_ok),
-    Route("/{whatever:path}", pipeline_route, methods=["POST"]),
-]
-
-middleware = [Middleware(GZipMiddleware, minimum_size=1000)]
-if os.environ.get("DEBUG", "") == "1":
-    from starlette.middleware.cors import CORSMiddleware
-
-    middleware.append(
-        Middleware(
-            CORSMiddleware,
-            allow_origins=["*"],
-            allow_headers=["*"],
-            allow_methods=["*"],
-        )
-    )
-
-app = Starlette(routes=routes, middleware=middleware)
-
-
-@app.on_event("startup")
-async def startup_event():
-    logger = logging.getLogger("uvicorn.access")
-    handler = logging.StreamHandler()
-    handler.setFormatter(logging.Formatter("%(asctime)s - %(levelname)s - %(message)s"))
-    logger.handlers = [handler]
-
-    # Link between `api-inference-community` and framework code.
-    app.get_pipeline = get_pipeline
-    try:
-        get_pipeline()
-    except Exception:
-        # We can fail so we can show exception later.
-        pass
-
-
-if __name__ == "__main__":
-    try:
-        get_pipeline()
-    except Exception:
-        # We can fail so we can show exception later.
-        pass
diff --git a/docker_images/superb/app/pipelines/__init__.py b/docker_images/superb/app/pipelines/__init__.py
deleted file mode 100644
index 8e8c4c8f..00000000
--- a/docker_images/superb/app/pipelines/__init__.py
+++ /dev/null
@@ -1,6 +0,0 @@
-from app.pipelines.base import Pipeline, PipelineException  # isort:skip
-
-from app.pipelines.automatic_speech_recognition import (
-    AutomaticSpeechRecognitionPipeline,
-)
-from app.pipelines.speech_segmentation import SpeechSegmentationPipeline
diff --git a/docker_images/superb/app/pipelines/automatic_speech_recognition.py b/docker_images/superb/app/pipelines/automatic_speech_recognition.py
deleted file mode 100644
index fcabf3fb..00000000
--- a/docker_images/superb/app/pipelines/automatic_speech_recognition.py
+++ /dev/null
@@ -1,49 +0,0 @@
-import os
-import subprocess
-import sys
-from typing import Dict
-
-import numpy as np
-from app.pipelines import Pipeline
-from huggingface_hub import snapshot_download
-
-
-class AutomaticSpeechRecognitionPipeline(Pipeline):
-    def __init__(self, model_id: str):
-        # IMPLEMENT_THIS
-        # Preload all the elements you are going to need at inference.
-        # For instance your model, processors, tokenizer that might be needed.
-        # This function is only called once, so do all the heavy processing I/O here
-        # IMPLEMENT_THIS : Please define a `self.sampling_rate` for this pipeline
-        # to automatically read the input correctly
-        filepath = snapshot_download(model_id)
-        sys.path.append(filepath)
-        if "requirements.txt" in os.listdir(filepath):
-            subprocess.check_call(
-                [
-                    sys.executable,
-                    "-m",
-                    "pip",
-                    "install",
-                    "-r",
-                    os.path.join(filepath, "requirements.txt"),
-                ]
-            )
-
-        from model import PreTrainedModel
-
-        self.model = PreTrainedModel(filepath)
-        self.sampling_rate = 16000
-
-    def __call__(self, inputs: np.array) -> Dict[str, str]:
-        """
-        Args:
-            inputs (:obj:`np.array`):
-                The raw waveform of audio received. By default at 16KHz.
-                Check `app.validation` if a different sample rate is required
-                or if it depends on the model
-        Return:
-            A :obj:`dict`:. The object return should be liked {"text": "XXX"} containing
-            the detected langage from the input audio
-        """
-        return self.model(inputs)
diff --git a/docker_images/superb/app/pipelines/base.py b/docker_images/superb/app/pipelines/base.py
deleted file mode 100644
index 2d846eb4..00000000
--- a/docker_images/superb/app/pipelines/base.py
+++ /dev/null
@@ -1,16 +0,0 @@
-from abc import ABC, abstractmethod
-from typing import Any
-
-
-class Pipeline(ABC):
-    @abstractmethod
-    def __init__(self, model_id: str):
-        raise NotImplementedError("Pipelines should implement an __init__ method")
-
-    @abstractmethod
-    def __call__(self, inputs: Any) -> Any:
-        raise NotImplementedError("Pipelines should implement a __call__ method")
-
-
-class PipelineException(Exception):
-    pass
diff --git a/docker_images/superb/app/pipelines/speech_segmentation.py b/docker_images/superb/app/pipelines/speech_segmentation.py
deleted file mode 100644
index ac7f172b..00000000
--- a/docker_images/superb/app/pipelines/speech_segmentation.py
+++ /dev/null
@@ -1,56 +0,0 @@
-import os
-import subprocess
-import sys
-from typing import Dict, List, Union
-
-import numpy as np
-from api_inference_community.normalizers import speaker_diarization_normalize
-from app.pipelines import Pipeline
-from huggingface_hub import snapshot_download
-
-
-class SpeechSegmentationPipeline(Pipeline):
-    def __init__(self, model_id: str):
-        # IMPLEMENT_THIS
-        # Preload all the elements you are going to need at inference.
-        # For instance your model, processors, tokenizer that might be needed.
-        # This function is only called once, so do all the heavy processing I/O here
-        # IMPLEMENT_THIS : Please define a `self.sampling_rate` for this pipeline
-        # to automatically read the input correctly
-        filepath = snapshot_download(model_id)
-        sys.path.append(filepath)
-        if "requirements.txt" in os.listdir(filepath):
-            subprocess.check_call(
-                [
-                    sys.executable,
-                    "-m",
-                    "pip",
-                    "install",
-                    "-r",
-                    os.path.join(filepath, "requirements.txt"),
-                ]
-            )
-
-        from model import PreTrainedModel
-
-        self.model = PreTrainedModel(filepath)
-        self.sampling_rate = 16000
-
-    def __call__(self, inputs: np.array) -> List[Dict[str, Union[str, float]]]:
-        """
-        Args:
-            inputs (:obj:`np.array`):
-                The raw waveform of audio received. By default at self.sampling_rate, otherwise 16KHz.
-        Return:
-            A :obj:`list`:. Each item in the list is like {"class": "XXX", "start": float, "end": float}
-            "class" is the associated class of the audio segment, "start" and "end" are markers expressed in seconds
-            within the audio file. Segments can overlap in any way the want.
-        """
-        # S x N boolean tensor
-        # S : sequence_length
-        # N : Number of expected speakers
-        # Filled with ones where speaker-n is speaking
-        outputs = self.model(inputs)
-        return speaker_diarization_normalize(
-            outputs, self.sampling_rate, ["SPEAKER_0", "SPEAKER_1"]
-        )
diff --git a/docker_images/superb/prestart.sh b/docker_images/superb/prestart.sh
deleted file mode 100644
index 67cbe056..00000000
--- a/docker_images/superb/prestart.sh
+++ /dev/null
@@ -1,2 +0,0 @@
-git-lfs install
-python app/main.py
diff --git a/docker_images/superb/requirements.txt b/docker_images/superb/requirements.txt
deleted file mode 100644
index b9a83140..00000000
--- a/docker_images/superb/requirements.txt
+++ /dev/null
@@ -1,6 +0,0 @@
-starlette==0.14.2
-api-inference-community==0.0.23
-git+https://github.com/pytorch/fairseq.git@f2146bd#egg=fairseq
-git+https://github.com/huggingface/s3prl.git@huggingface2#egg=s3prl
-hf-lfs
-huggingface_hub==0.5.1
diff --git a/docker_images/superb/tests/__init__.py b/docker_images/superb/tests/__init__.py
deleted file mode 100644
index e69de29b..00000000
diff --git a/docker_images/superb/tests/samples/malformed.flac b/docker_images/superb/tests/samples/malformed.flac
deleted file mode 100644
index 06d74050..00000000
Binary files a/docker_images/superb/tests/samples/malformed.flac and /dev/null differ
diff --git a/docker_images/superb/tests/samples/plane.jpg b/docker_images/superb/tests/samples/plane.jpg
deleted file mode 100644
index 9e0cc585..00000000
Binary files a/docker_images/superb/tests/samples/plane.jpg and /dev/null differ
diff --git a/docker_images/superb/tests/samples/plane2.jpg b/docker_images/superb/tests/samples/plane2.jpg
deleted file mode 100644
index b1f076e3..00000000
Binary files a/docker_images/superb/tests/samples/plane2.jpg and /dev/null differ
diff --git a/docker_images/superb/tests/samples/sample1.flac b/docker_images/superb/tests/samples/sample1.flac
deleted file mode 100644
index 837cd984..00000000
Binary files a/docker_images/superb/tests/samples/sample1.flac and /dev/null differ
diff --git a/docker_images/superb/tests/samples/sample1.webm b/docker_images/superb/tests/samples/sample1.webm
deleted file mode 100644
index 7c449a27..00000000
Binary files a/docker_images/superb/tests/samples/sample1.webm and /dev/null differ
diff --git a/docker_images/superb/tests/samples/sample1_dual.ogg b/docker_images/superb/tests/samples/sample1_dual.ogg
deleted file mode 100644
index 1e8524e0..00000000
Binary files a/docker_images/superb/tests/samples/sample1_dual.ogg and /dev/null differ
diff --git a/docker_images/superb/tests/test_api.py b/docker_images/superb/tests/test_api.py
deleted file mode 100644
index ceb691ac..00000000
--- a/docker_images/superb/tests/test_api.py
+++ /dev/null
@@ -1,53 +0,0 @@
-import os
-from typing import Dict, List
-from unittest import TestCase, skipIf
-
-from app.main import ALLOWED_TASKS, get_pipeline
-
-
-# Must contain at least one example of each implemented pipeline
-# Tests do not check the actual values of the model output, so small dummy
-# models are recommended for faster tests.
-TESTABLE_MODELS: Dict[str, List[str]] = {
-    "automatic-speech-recognition": [
-        # Repo using transformers library.
-        "osanseviero/asr-with-transformers-wav2vec2",
-        # Repo using s3prl.
-        "osanseviero/hubert_s3prl",
-        # Repo using s3prl and installing a custom dependency.
-        "osanseviero/hubert_s3prl_req",
-    ],
-    "speech-segmentation": "osanseviero/hubert-sd",
-}
-
-
-ALL_TASKS = {
-    "automatic-speech-recognition",
-    "feature-extraction",
-    "image-classification",
-    "question-answering",
-    "sentence-similarity",
-    "text-generation",
-    "text-to-speech",
-    "token-classification",
-}
-
-
-class PipelineTestCase(TestCase):
-    @skipIf(
-        os.path.dirname(os.path.dirname(__file__)).endswith("common"),
-        "common is a special case",
-    )
-    def test_has_at_least_one_task_enabled(self):
-        self.assertGreater(
-            len(ALLOWED_TASKS.keys()), 0, "You need to implement at least one task"
-        )
-
-    def test_unsupported_tasks(self):
-        unsupported_tasks = ALL_TASKS - ALLOWED_TASKS.keys()
-        for unsupported_task in unsupported_tasks:
-            with self.subTest(msg=unsupported_task, task=unsupported_task):
-                os.environ["TASK"] = unsupported_task
-                os.environ["MODEL_ID"] = "XX"
-                with self.assertRaises(EnvironmentError):
-                    get_pipeline()
diff --git a/docker_images/superb/tests/test_api_automatic_speech_recognition.py b/docker_images/superb/tests/test_api_automatic_speech_recognition.py
deleted file mode 100644
index 941bec2f..00000000
--- a/docker_images/superb/tests/test_api_automatic_speech_recognition.py
+++ /dev/null
@@ -1,100 +0,0 @@
-import json
-import os
-from unittest import TestCase, skipIf
-
-from app.main import ALLOWED_TASKS
-from parameterized import parameterized_class
-from starlette.testclient import TestClient
-from tests.test_api import TESTABLE_MODELS
-
-
-@skipIf(
-    "automatic-speech-recognition" not in ALLOWED_TASKS,
-    "automatic-speech-recognition not implemented",
-)
-@parameterized_class(
-    [
-        {"model_id": model_id}
-        for model_id in TESTABLE_MODELS["automatic-speech-recognition"]
-    ]
-)
-class AutomaticSpeecRecognitionTestCase(TestCase):
-    def setUp(self):
-        self.old_model_id = os.getenv("MODEL_ID")
-        self.old_task = os.getenv("TASK")
-        os.environ["MODEL_ID"] = self.model_id
-        os.environ["TASK"] = "automatic-speech-recognition"
-
-        from app.main import app, get_pipeline
-
-        get_pipeline.cache_clear()
-
-        self.app = app
-
-    def tearDown(self):
-        if self.old_model_id is not None:
-            os.environ["MODEL_ID"] = self.old_model_id
-        else:
-            del os.environ["MODEL_ID"]
-        if self.old_task is not None:
-            os.environ["TASK"] = self.old_task
-        else:
-            del os.environ["TASK"]
-
-    def read(self, filename: str) -> bytes:
-        dirname = os.path.dirname(os.path.abspath(__file__))
-        filename = os.path.join(dirname, "samples", filename)
-        with open(filename, "rb") as f:
-            bpayload = f.read()
-        return bpayload
-
-    def test_original_audiofile(self):
-        bpayload = self.read("sample1.flac")
-
-        with TestClient(self.app) as client:
-            response = client.post("/", data=bpayload)
-
-        self.assertEqual(
-            response.status_code,
-            200,
-        )
-        content = json.loads(response.content)
-        self.assertEqual(set(content.keys()), {"text"})
-
-    def test_malformed_audio(self):
-        bpayload = self.read("malformed.flac")
-
-        with TestClient(self.app) as client:
-            response = client.post("/", data=bpayload)
-
-        self.assertEqual(
-            response.status_code,
-            400,
-        )
-        self.assertEqual(response.content, b'{"error":"Malformed soundfile"}')
-
-    def test_dual_channel_audiofile(self):
-        bpayload = self.read("sample1_dual.ogg")
-
-        with TestClient(self.app) as client:
-            response = client.post("/", data=bpayload)
-
-        self.assertEqual(
-            response.status_code,
-            200,
-        )
-        content = json.loads(response.content)
-        self.assertEqual(set(content.keys()), {"text"})
-
-    def test_webm_audiofile(self):
-        bpayload = self.read("sample1.webm")
-
-        with TestClient(self.app) as client:
-            response = client.post("/", data=bpayload)
-
-        self.assertEqual(
-            response.status_code,
-            200,
-        )
-        content = json.loads(response.content)
-        self.assertEqual(set(content.keys()), {"text"})
diff --git a/docker_images/superb/tests/test_api_speech_segmentation.py b/docker_images/superb/tests/test_api_speech_segmentation.py
deleted file mode 100644
index c88292f1..00000000
--- a/docker_images/superb/tests/test_api_speech_segmentation.py
+++ /dev/null
@@ -1,112 +0,0 @@
-import json
-import os
-from unittest import TestCase, skipIf
-
-from app.main import ALLOWED_TASKS
-from starlette.testclient import TestClient
-from tests.test_api import TESTABLE_MODELS
-
-
-@skipIf(
-    "speech-segmentation" not in ALLOWED_TASKS,
-    "speech-segmentation not implemented",
-)
-class SpeechSegmentationTestCase(TestCase):
-    def setUp(self):
-        model_id = TESTABLE_MODELS["speech-segmentation"]
-        self.old_model_id = os.getenv("MODEL_ID")
-        self.old_task = os.getenv("TASK")
-        os.environ["MODEL_ID"] = model_id
-        os.environ["TASK"] = "speech-segmentation"
-        from app.main import app
-
-        self.app = app
-
-    @classmethod
-    def setUpClass(cls):
-        from app.main import get_pipeline
-
-        get_pipeline.cache_clear()
-
-    def tearDown(self):
-        if self.old_model_id is not None:
-            os.environ["MODEL_ID"] = self.old_model_id
-        else:
-            del os.environ["MODEL_ID"]
-        if self.old_task is not None:
-            os.environ["TASK"] = self.old_task
-        else:
-            del os.environ["TASK"]
-
-    def read(self, filename: str) -> bytes:
-        dirname = os.path.dirname(os.path.abspath(__file__))
-        filename = os.path.join(dirname, "samples", filename)
-        with open(filename, "rb") as f:
-            bpayload = f.read()
-        return bpayload
-
-    def test_original_audiofile(self):
-        bpayload = self.read("sample1.flac")
-
-        with TestClient(self.app) as client:
-            response = client.post("/", data=bpayload)
-
-        self.assertEqual(
-            response.status_code,
-            200,
-        )
-        content = json.loads(response.content)
-        self.assertIsInstance(content, list)
-        for c in content:
-            self.assertEqual(set(c.keys()), {"class", "start", "end"})
-            self.assertIsInstance(c["class"], str)
-            self.assertIsInstance(c["start"], float)
-            self.assertIsInstance(c["end"], float)
-
-    def test_malformed_audio(self):
-        bpayload = self.read("malformed.flac")
-
-        with TestClient(self.app) as client:
-            response = client.post("/", data=bpayload)
-
-        self.assertEqual(
-            response.status_code,
-            400,
-        )
-        self.assertEqual(response.content, b'{"error":"Malformed soundfile"}')
-
-    def test_dual_channel_audiofile(self):
-        bpayload = self.read("sample1_dual.ogg")
-
-        with TestClient(self.app) as client:
-            response = client.post("/", data=bpayload)
-
-        self.assertEqual(
-            response.status_code,
-            200,
-        )
-        content = json.loads(response.content)
-        self.assertIsInstance(content, list)
-        for c in content:
-            self.assertEqual(set(c.keys()), {"class", "start", "end"})
-            self.assertIsInstance(c["class"], str)
-            self.assertIsInstance(c["start"], float)
-            self.assertIsInstance(c["end"], float)
-
-    def test_webm_audiofile(self):
-        bpayload = self.read("sample1.webm")
-
-        with TestClient(self.app) as client:
-            response = client.post("/", data=bpayload)
-
-        self.assertEqual(
-            response.status_code,
-            200,
-        )
-        content = json.loads(response.content)
-        self.assertIsInstance(content, list)
-        for c in content:
-            self.assertEqual(set(c.keys()), {"class", "start", "end"})
-            self.assertIsInstance(c["class"], str)
-            self.assertIsInstance(c["start"], float)
-            self.assertIsInstance(c["end"], float)
diff --git a/docker_images/superb/tests/test_docker_build.py b/docker_images/superb/tests/test_docker_build.py
deleted file mode 100644
index ca5dd65d..00000000
--- a/docker_images/superb/tests/test_docker_build.py
+++ /dev/null
@@ -1,23 +0,0 @@
-import os
-import subprocess
-from unittest import TestCase
-
-
-class cd:
-    """Context manager for changing the current working directory"""
-
-    def __init__(self, newPath):
-        self.newPath = os.path.expanduser(newPath)
-
-    def __enter__(self):
-        self.savedPath = os.getcwd()
-        os.chdir(self.newPath)
-
-    def __exit__(self, etype, value, traceback):
-        os.chdir(self.savedPath)
-
-
-class DockerBuildTestCase(TestCase):
-    def test_can_build_docker_image(self):
-        with cd(os.path.dirname(os.path.dirname(__file__))):
-            subprocess.check_output(["docker", "build", "."])
diff --git a/tests/test_dockers.py b/tests/test_dockers.py
index 2ac43f17..4f726fa5 100644
--- a/tests/test_dockers.py
+++ b/tests/test_dockers.py
@@ -327,33 +327,6 @@ def test_doctr(self):
         )
         self.framework_invalid_test("doctr")
 
-    def test_superb(self):
-        # Very basic repo just using transformers.
-        self.framework_docker_test(
-            "superb",
-            "automatic-speech-recognition",
-            "osanseviero/asr-with-transformers-wav2vec2",
-        )
-        self.framework_docker_test(
-            "superb",
-            "speech-segmentation",
-            "osanseviero/hubert-sd",
-        )
-        # # Too slow, requires downloading the upstream model from PyTorch Hub which is quite heavy
-        # # self.framework_docker_test(
-        # #    "superb", "automatic-speech-recognition", "osanseviero/hubert_s3prl_req"
-        # # )
-        self.framework_invalid_test("superb")
-        self.framework_docker_batch(
-            "superb",
-            "automatic-speech-recognition",
-            "osanseviero/asr-with-transformers-wav2vec2",
-            dataset_name="Narsil/asr_dummy",
-            dataset_config="asr",
-            dataset_split="test",
-            dataset_column="file",
-        )
-
     def test_nemo(self):
         self.framework_docker_test(
             "nemo", "automatic-speech-recognition", "nvidia/stt_en_conformer_ctc_medium"