improve doc around supported tasks and accelertor options

Author: rbrugaro

optimum/intel/pipelines/pipeline_base.py

@@ -171,6 +171,12 @@ def pipeline(
             The task defining which pipeline will be returned. Currently accepted tasks are:
 
             - `"text-generation"`: will return a [`TextGenerationPipeline`]:.
+            - `"fill-mask"`: will return a [`FillMaskPipeline`].
+            - `"question-answering"`: will return a [`QuestionAnsweringPipeline`].
+            - `"image-classificatio"`: will return a [`ImageClassificationPipeline`].
+            - `"text-classification"`: will return a [`TextClassificationPipeline`].
+            - `"token-classification"`: will return a [`TokenClassificationPipeline`].
+            - `"audio-classification"`: will return a [`AudioClassificationPipeline`].
 
         model (`str` or [`PreTrainedModel`], *optional*):
             The model that will be used by the pipeline to make predictions. This can be a model identifier or an
@@ -185,8 +191,8 @@ def pipeline(
             is not specified or not a string, then the default tokenizer for `config` is loaded (if it is a string).
             However, if `config` is also not given or not a string, then the default tokenizer for the given `task`
             will be loaded.
-        accelerator (`str`, *optional*, defaults to `"ipex"`):
-            The optimization backends, choose from ["ipex", "inc", "openvino"].
+        accelerator (`str`, *optional*):
+            The optimization backends, choose from ["ipex"].
         use_fast (`bool`, *optional*, defaults to `True`):
             Whether or not to use a Fast tokenizer if possible (a [`PreTrainedTokenizerFast`]).
         torch_dtype (`str` or `torch.dtype`, *optional*):