Fix and document Zero Shot Image Classification (#16079)

docs/source/main_classes/pipelines.mdx

@@ -39,6 +39,7 @@ There are two categories of pipeline abstractions to be aware about:
   - [`TokenClassificationPipeline`]
   - [`TranslationPipeline`]
   - [`ZeroShotClassificationPipeline`]
+  - [`ZeroShotImageClassificationPipeline`]
 
 ## The pipeline abstraction
 

src/transformers/pipelines/__init__.py

@@ -245,7 +245,7 @@ SUPPORTED_TASKS = {
         "impl": ZeroShotImageClassificationPipeline,
         "tf": (TFAutoModel,) if is_tf_available() else (),
         "pt": (AutoModel,) if is_torch_available() else (),
-        "default": {"pt": "openai/clip-vit-base-patch32", "tf": "openai/clip-vit-base-patch32"},
+        "default": {"model": {"pt": "openai/clip-vit-base-patch32", "tf": "openai/clip-vit-base-patch32"}},
         "type": "multimodal",
     },
     "conversational": {
@@ -346,6 +346,7 @@ def check_task(task: str) -> Tuple[Dict, Any]:
             - `"translation_xx_to_yy"`
             - `"summarization"`
             - `"zero-shot-classification"`
+            - `"zero-shot-image-classification"`
 
     Returns:
         (task_defaults`dict`, task_options: (`tuple`, None)) The actual dictionary required to initialize the pipeline

src/transformers/pipelines/zero_shot_image_classification.py

@@ -35,7 +35,7 @@ class ZeroShotImageClassificationPipeline(ChunkPipeline):
     `"zero-shot-image-classification"`.
 
     See the list of available models on
-    [huggingface.co/models](https://huggingface.co/models?filter=zer-shot-image-classification).
+    [huggingface.co/models](https://huggingface.co/models?filter=zero-shot-image-classification).
     """
 
     def __init__(self, **kwargs):