[Doc]: fix typos in various files (#24798)
Signed-off-by: Didier Durand <durand.didier@gmail.com>
This commit is contained in:
@@ -1446,7 +1446,7 @@ class Ernie4_5_VLMoeForConditionalGeneration(nn.Module, SupportsMultiModal,
|
||||
return None
|
||||
|
||||
# The result multimodal_embeddings is tuple of tensors, with each
|
||||
# tensor correspoending to a multimodal data item (image or video).
|
||||
# tensor corresponding to a multimodal data item (image or video).
|
||||
multimodal_embeddings: tuple[torch.Tensor, ...] = ()
|
||||
|
||||
# NOTE: It is important to iterate over the keys in this dictionary
|
||||
|
||||
@@ -586,10 +586,10 @@ class Gemma3nForConditionalGeneration(nn.Module, SupportsMultiModal,
|
||||
|
||||
# ruff: noqa
|
||||
# The Gemma3nProcessor expects all audio will be 30s in length and inserts 188 audio soft tokens into the
|
||||
# text to account for this. However, the audio preprocessing and encoder do not gurarantee they will
|
||||
# text to account for this. However, the audio preprocessing and encoder do not guarantee they will
|
||||
# produce 188 soft tokens; they will produce at most that many tokens, but they may produce fewer tokens
|
||||
# depending on the length of the longest audio input in the batch. When we encounter this situation, we pad
|
||||
# the audio feature out to 188 soft tokens with the emebedding of the last token in the embed_audio vocab.
|
||||
# the audio feature out to 188 soft tokens with the embedding of the last token in the embed_audio vocab.
|
||||
# TODO precompute and cache padding
|
||||
audio_padding_toks = torch.tensor([[self.vocab_size - 1]],
|
||||
dtype=torch.long,
|
||||
|
||||
@@ -560,7 +560,7 @@ class LlamaNemotronVLChatModel(nn.Module, SupportsMultiModal, SupportsPP,
|
||||
return []
|
||||
|
||||
# The result multimodal_embeddings is tuple of tensors, with each
|
||||
# tensor correspoending to a multimodal data item (image).
|
||||
# tensor corresponding to a multimodal data item (image).
|
||||
multimodal_embeddings: tuple[torch.Tensor, ...] = ()
|
||||
|
||||
# NOTE: It is important to iterate over the keys in this dictionary
|
||||
|
||||
@@ -1154,7 +1154,7 @@ class Phi4MMForCausalLM(nn.Module, SupportsLoRA, SupportsMultiModal):
|
||||
return None
|
||||
|
||||
# The result multimodal_embeddings is tuple of tensors, with each
|
||||
# tensor correspoending to a multimodal data item (image or video).
|
||||
# tensor corresponding to a multimodal data item (image or video).
|
||||
multimodal_embeddings: tuple[torch.Tensor, ...] = ()
|
||||
|
||||
# NOTE: It is important to iterate over the keys in this dictionary
|
||||
|
||||
Reference in New Issue
Block a user