Add full API docs and improve the UX of navigating them (#17485)

Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>

tests/conftest.py

@@ -112,11 +112,11 @@ class AudioTestAssets(list[AudioAsset]):
 
 
 IMAGE_ASSETS = ImageTestAssets()
-"""Singleton instance of :class:`ImageTestAssets`."""
+"""Singleton instance of {class}`ImageTestAssets`."""
 VIDEO_ASSETS = VideoTestAssets()
-"""Singleton instance of :class:`VideoTestAssets`."""
+"""Singleton instance of {class}`VideoTestAssets`."""
 AUDIO_ASSETS = AudioTestAssets()
-"""Singleton instance of :class:`AudioTestAssets`."""
+"""Singleton instance of {class}`AudioTestAssets`."""
 
 
 @pytest.fixture(scope="function", autouse=True)
@@ -724,7 +724,7 @@ def hf_runner():
 class VllmRunner:
     """
     The default value of some arguments have been modified from
-    :class:`~vllm.LLM` as follows:
+    {class}`~vllm.LLM` as follows:
 
     - `trust_remote_code`: Set to `True` instead of `False` for convenience.
     - `seed`: Set to `0` instead of `None` for test reproducibility.

tests/tokenization/test_get_eos.py

@@ -2,7 +2,7 @@
 """
 This test file includes some cases where it is inappropriate to
 only get the `eos_token_id` from the tokenizer as defined by
-:meth:`vllm.LLMEngine._get_eos_token_id`.
+{meth}`vllm.LLMEngine._get_eos_token_id`.
 """
 from vllm.transformers_utils.config import try_get_generation_config
 from vllm.transformers_utils.tokenizer import get_tokenizer

tests/utils.py

@@ -952,7 +952,7 @@ def get_client_text_logprob_generations(
         completions: list[Completion]) -> list[TextTextLogprobs]:
     '''Operates on the output of a request made to an Open-AI-protocol
     completions endpoint; obtains top-rank logprobs for each token in
-    each :class:`SequenceGroup`
+    each {class}`SequenceGroup`
     '''
     text_generations = get_client_text_generations(completions)
     text = ''.join(text_generations)

tests/v1/core/test_scheduler.py

@@ -44,7 +44,7 @@ def create_scheduler(
                              (None)
 
     Returns:
-      :class:`Scheduler` instance
+      {class}`Scheduler` instance
     '''
     if max_model_len is None:
         max_model_len = max_num_batched_tokens