mirror of
https://github.com/huggingface/text-generation-inference.git
synced 2025-04-22 15:32:08 +00:00
433934519c
Superseeds #1459 The fix works as follows. We updated next_token_chooser to return all logprbs, then batch_top_n_tokens, now also gets accepted_ids + speculated_length (so it knows how to interpret the flat logprobs). We then update the code to return lists ot `Tokens` that it expects. <!-- Congratulations! You've made it this far! You're not quite done yet though. Once merged, your PR is going to appear in the release notes with the title you set, so make sure it's a great title that fully reflects the extent of your awesome contribution. Then, please replace this with a description of the change and which issue is fixed (if applicable). Please also include relevant motivation and context. List any dependencies (if any) that are required for this change. Once you're done, someone will review your PR shortly (see the section "Who can review?" below to tag some potential reviewers). They may suggest changes to make the code even better. If no one reviewed your PR after a week has passed, don't hesitate to post a new comment @-mentioning the same persons---sometimes notifications get lost. --> <!-- Remove if not applicable --> Fixes # (issue) - [ ] This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case). - [ ] Did you read the [contributor guideline](https://github.com/huggingface/transformers/blob/main/CONTRIBUTING.md#start-contributing-pull-requests), Pull Request section? - [ ] Was this discussed/approved via a Github issue or the [forum](https://discuss.huggingface.co/)? Please add a link to it if that's the case. - [ ] Did you make sure to update the documentation with your changes? Here are the [documentation guidelines](https://github.com/huggingface/transformers/tree/main/docs), and [here are tips on formatting docstrings](https://github.com/huggingface/transformers/tree/main/docs#writing-source-documentation). - [ ] Did you write any new necessary tests? Anyone in the community is free to review the PR once the tests have passed. Feel free to tag members/contributors who may be interested in your PR. <!-- Your PR will be replied to more quickly if you can figure out the right person to tag with @ @OlivierDehaene OR @Narsil -->
133 lines
4.4 KiB
Python
133 lines
4.4 KiB
Python
# Copyright (C) 2024 Habana Labs, Ltd. an Intel Company.
|
|
|
|
import pytest
|
|
import torch
|
|
from transformers import AutoTokenizer
|
|
|
|
from text_generation_server.utils.tokens import (
|
|
StopSequenceCriteria,
|
|
StoppingCriteria,
|
|
FinishReason,
|
|
batch_top_tokens,
|
|
make_tokenizer_optional,
|
|
)
|
|
|
|
|
|
@pytest.fixture
|
|
def skip_tokenizer_env_var():
|
|
import os
|
|
os.environ["SKIP_TOKENIZER_IN_TGI"] = "true"
|
|
yield
|
|
del os.environ['SKIP_TOKENIZER_IN_TGI']
|
|
|
|
|
|
def test_stop_sequence_criteria():
|
|
criteria = StopSequenceCriteria("/test;")
|
|
|
|
assert not criteria("/")
|
|
assert not criteria("/test")
|
|
assert criteria("/test;")
|
|
assert not criteria("/test; ")
|
|
|
|
|
|
def test_stop_sequence_criteria_escape():
|
|
criteria = StopSequenceCriteria("<|stop|>")
|
|
|
|
assert not criteria("<")
|
|
assert not criteria("<|stop")
|
|
assert criteria("<|stop|>")
|
|
assert not criteria("<|stop|> ")
|
|
|
|
|
|
def test_stopping_criteria():
|
|
criteria = StoppingCriteria(0, [StopSequenceCriteria("/test;")], max_new_tokens=5)
|
|
assert criteria(65827, "/test") == (False, None)
|
|
assert criteria(30, ";") == (True, FinishReason.FINISH_REASON_STOP_SEQUENCE)
|
|
|
|
|
|
def test_stopping_criteria_eos():
|
|
criteria = StoppingCriteria(0, [StopSequenceCriteria("/test;")], max_new_tokens=5)
|
|
assert criteria(1, "") == (False, None)
|
|
assert criteria(0, "") == (True, FinishReason.FINISH_REASON_EOS_TOKEN)
|
|
|
|
|
|
def test_stopping_criteria_max():
|
|
criteria = StoppingCriteria(0, [StopSequenceCriteria("/test;")], max_new_tokens=5)
|
|
assert criteria(1, "") == (False, None)
|
|
assert criteria(1, "") == (False, None)
|
|
assert criteria(1, "") == (False, None)
|
|
assert criteria(1, "") == (False, None)
|
|
assert criteria(1, "") == (True, FinishReason.FINISH_REASON_LENGTH)
|
|
|
|
|
|
def test_batch_top_tokens():
|
|
top_n_tokens = [0, 2, 3, 4, 5]
|
|
top_n_tokens_tensor = torch.tensor(top_n_tokens)
|
|
inp_logprobs = torch.tensor([[-1.0, -3.0, -4.0, -2.0, -3.0]] * 5)
|
|
accepted_ids = torch.ones_like(top_n_tokens_tensor)
|
|
|
|
topn_tok_ids, topn_tok_logprobs = batch_top_tokens(
|
|
top_n_tokens, top_n_tokens_tensor, inp_logprobs, accepted_ids
|
|
)
|
|
|
|
assert topn_tok_ids[0] == [[]]
|
|
assert topn_tok_ids[1] == [[0, 3]]
|
|
assert topn_tok_ids[2] == [[0, 3, 1, 4]]
|
|
assert topn_tok_ids[3] == [[0, 3, 1, 4]]
|
|
assert topn_tok_ids[4] == [[0, 3, 1, 4, 2]]
|
|
|
|
assert topn_tok_logprobs[0] == [[]]
|
|
assert topn_tok_logprobs[1] == [[-1, -2]]
|
|
assert topn_tok_logprobs[2] == [[-1, -2, -3, -3]]
|
|
assert topn_tok_logprobs[3] == [[-1, -2, -3, -3]]
|
|
assert topn_tok_logprobs[4] == [[-1, -2, -3, -3, -4]]
|
|
|
|
# Now let's make second member of the batch be speculated
|
|
inp_logprobs = torch.tensor([[-1.0, -3.0, -4.0, -2.0, -3.0]] * 5 * 2)
|
|
accepted_ids[1] = 2
|
|
topn_tok_ids, topn_tok_logprobs = batch_top_tokens(
|
|
top_n_tokens, top_n_tokens_tensor, inp_logprobs, accepted_ids
|
|
)
|
|
|
|
assert topn_tok_ids[0] == [[]]
|
|
assert topn_tok_ids[1] == [[0, 3], [0, 3]]
|
|
assert topn_tok_ids[2] == [[0, 3, 1, 4]]
|
|
assert topn_tok_ids[3] == [[0, 3, 1, 4]]
|
|
assert topn_tok_ids[4] == [[0, 3, 1, 4, 2]]
|
|
|
|
assert topn_tok_logprobs[0] == [[]]
|
|
assert topn_tok_logprobs[1] == [[-1, -2], [-1, -2]]
|
|
assert topn_tok_logprobs[2] == [[-1, -2, -3, -3]]
|
|
assert topn_tok_logprobs[3] == [[-1, -2, -3, -3]]
|
|
assert topn_tok_logprobs[4] == [[-1, -2, -3, -3, -4]]
|
|
|
|
|
|
def test_pass_through_tokenizer(skip_tokenizer_env_var):
|
|
tokenizer = AutoTokenizer.from_pretrained(
|
|
'meta-llama/Llama-2-7b-chat-hf',
|
|
revision=None,
|
|
padding_side="left",
|
|
truncation_side="left",
|
|
)
|
|
tokenizer.pad_token_id = 2
|
|
make_tokenizer_optional(tokenizer)
|
|
|
|
input = ["1, 1724, 338, 6483, 6509, 29973", "?"]
|
|
tokenized_inputs = tokenizer(
|
|
input,
|
|
return_tensors="pt",
|
|
padding="max_length",
|
|
return_token_type_ids=False,
|
|
truncation=True,
|
|
max_length=1024,
|
|
)
|
|
assert tokenized_inputs['input_ids'].size() == torch.Size([2, 1024])
|
|
assert torch.equal(tokenized_inputs['input_ids'][0][1018:], torch.tensor([1, 1724, 338, 6483, 6509, 29973]))
|
|
assert torch.equal(tokenized_inputs['input_ids'][1][1023:], torch.tensor([tokenizer.pad_token_id]))
|
|
decoded_tokens = tokenizer.decode(tokenized_inputs["input_ids"][0], skip_special_tokens=True, clean_up_tokenization_spaces=False)
|
|
assert decoded_tokens.split(',')[1018:] == ['1', '1724', '338', '6483', '6509', '29973']
|
|
|
|
|
|
if __name__ == "__main__":
|
|
test_pass_through_tokenizer()
|