[Misc] Consolidate and optimize logic for building padded tensors #6541
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
DarkLight1337
commented
Jul 18, 2024
Comment on lines
-127
to
-132
| _STR_DTYPE_TO_TORCH_DTYPE = { | ||
| "half": torch.half, | ||
| "bfloat16": torch.bfloat16, | ||
| "float": torch.float, | ||
| } | ||
|
|
There was a problem hiding this comment.
I found that this mapping is a duplicate of the one in vllm.utils, so I've removed it.
| @@ -466,22 +465,30 @@ def from_lists(cls, temperatures: List[float], top_ps: List[float], | |||
| do_penalties = prompt_tokens or output_tokens | |||
|
|
|||
| if do_penalties: | |||
There was a problem hiding this comment.
I have merged the if-else blocks based on do_penalties together. Not sure why they were separated in the first place.
|
|
||
| The padding is applied to the end of each inner list until it reaches | ||
| `max_len`. | ||
| """ | ||
| padded_x = np.zeros([len(x), max_len], dtype=np.int32) + pad | ||
| padded_x = np.full((len(x), max_len), pad, dtype=dtype) |
There was a problem hiding this comment.
np.full(..., pad) is more efficient than np.zeros(...) + pad. Try it yourself:
python -m timeit "import numpy as np; np.zeros(100000) + 2"
python -m timeit "import numpy as np; np.full(100000, 2)"There was a problem hiding this comment.
I've also fixed the dtype to be consistent with the pytorch one.
DarkLight1337
changed the title
DarkLight1337
added
the
ready
DarkLight1337
commented
Jul 18, 2024
DarkLight1337
force-pushed
the
refactor-tensor-pad
branch
from
64382b3 to
38c5ab8
Compare
xjpang
pushed a commit
to xjpang/vllm
that referenced
this pull request
Jul 24, 2024
gnpinkert
pushed a commit
to gnpinkert/vllm
that referenced
this pull request
Jul 26, 2024
cduk
pushed a commit
to cduk/vllm-pascal
that referenced
this pull request
Aug 6, 2024
Alvant
pushed a commit
to compressa-ai/vllm
that referenced
this pull request
Oct 26, 2024
…lm-project#6541) Signed-off-by: Alvant <[email protected]>
KuntaiDu
pushed a commit
to KuntaiDu/vllm
that referenced
this pull request
Nov 20, 2024
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Following #6442 , this PR introduces a small refactor to clean up the code.
cc @peng1999