[test] Add LlamaDecoderLayerWithKVCache using captured input #215
Conversation
This file contains hidden or 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
It adds LlamaDecoderLayer test, which uses captured input. TICO-DCO-1.0-Signed-off-by: Sanggyu Lee <sg5.lee@samsung.com>
Contributor
Author
Contributor
Author
|
Hmm... (ADD) Ah, by default ./ccex format fixed the format error. I added the change. |
glistening
commented
Jul 17, 2025
Comment on lines
7
to
39
| import copy, inspect, types | ||
|
|
||
| from transformers.models.llama.modeling_llama import LlamaDecoderLayer | ||
|
|
||
| forward_old = LlamaDecoderLayer.forward | ||
|
|
||
|
|
||
| def capture_and_forward(self, *args, **kwargs): | ||
| global captured_input | ||
|
|
||
| # Prepare args tuple for TICO.convert() | ||
| # Get arg_names in positional args order using inspect | ||
| sig = inspect.signature(forward_old) | ||
| args_names = [ | ||
| # signature includes `self`` and `kwargs``. | ||
| # Just retrieve the ordinary positional inputs only | ||
| name for name in sig.parameters.keys() if name not in ("self", "kwargs") | ||
| ] | ||
|
|
||
| args_dict = dict(zip(args_names, args)) | ||
| args_dict.update(kwargs) | ||
|
|
||
| def populate_args(args_dict, filter): | ||
| for key in filter: | ||
| args_dict.pop(key, None) | ||
| args_tuple = tuple(args_dict.get(name, None) for name in args_names) | ||
| return copy.deepcopy(args_tuple) | ||
|
|
||
| if len(args_dict['past_key_value'].key_cache) != 0: | ||
| input_to_remove = [ "use_cache" ] | ||
| captured_input = populate_args(args_dict, input_to_remove) | ||
|
|
||
| return forward_old(self, *args, **kwargs) |
Contributor
Author
There was a problem hiding this comment.
It would be good to put this code in somewhere like TICO/util.
Comment on lines
42
to
55
| # Tokenizer | ||
| from transformers import AutoTokenizer | ||
|
|
||
| tokenizer = AutoTokenizer.from_pretrained(model_name) | ||
| tokenizer.pad_token = tokenizer.eos_token | ||
| tokenizer.padding_side = "right" | ||
| inputs = tokenizer( | ||
| prompt, | ||
| return_tensors="pt", | ||
| padding="max_length", | ||
| max_length=32, | ||
| truncation=True, | ||
| ) | ||
|
|
Contributor
Author
There was a problem hiding this comment.
I think this code can be generated simply using gpt for new models.
glistening
changed the title
glistening
force-pushed
the
decoderlayer
branch
4 times, most recently
from
99e8496 to
040cca8
Compare
Contributor
Author
|
See #217. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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.
It adds LlamaDecoderLayerWithKVCache test, which uses captured input.
TICO-DCO-1.0-Signed-off-by: Sanggyu Lee sg5.lee@samsung.com