hf_text-generation-inference/server/tests/models/test_santacoder.py

94 lines
2.9 KiB
Python
Raw Normal View History

2023-01-20 04:24:39 -07:00
import pytest
2023-03-07 10:52:22 -07:00
from text_generation_server.pb import generate_pb2
from text_generation_server.models.causal_lm import CausalLMBatch
from text_generation_server.models.santacoder import SantaCoder
2023-01-20 04:24:39 -07:00
@pytest.fixture(scope="session")
def default_santacoder():
return SantaCoder("bigcode/santacoder")
@pytest.fixture
def default_pb_request(default_pb_parameters, default_pb_stop_parameters):
return generate_pb2.Request(
id=0,
inputs="def",
truncate=100,
2023-01-20 04:24:39 -07:00
parameters=default_pb_parameters,
stopping_parameters=default_pb_stop_parameters,
)
@pytest.fixture
def default_pb_batch(default_pb_request):
return generate_pb2.Batch(id=0, requests=[default_pb_request], size=1)
@pytest.fixture
def default_fim_pb_request(default_pb_parameters, default_pb_stop_parameters):
return generate_pb2.Request(
id=0,
inputs="<fim-prefix>def<fim-suffix>world<fim-middle>",
truncate=100,
2023-01-20 04:24:39 -07:00
parameters=default_pb_parameters,
stopping_parameters=default_pb_stop_parameters,
)
@pytest.fixture
def default_fim_pb_batch(default_fim_pb_request):
return generate_pb2.Batch(id=0, requests=[default_fim_pb_request], size=1)
2023-01-20 07:35:22 -07:00
@pytest.mark.skip
2023-01-20 04:24:39 -07:00
def test_santacoder_generate_token_completion(default_santacoder, default_pb_batch):
batch = CausalLMBatch.from_pb(
default_pb_batch, default_santacoder.tokenizer, default_santacoder.device
)
next_batch = batch
for _ in range(batch.stopping_criterias[0].max_new_tokens - 1):
generations, next_batch = default_santacoder.generate_token(next_batch)
assert len(generations) == len(next_batch)
2023-01-20 04:24:39 -07:00
generations, next_batch = default_santacoder.generate_token(next_batch)
2023-01-20 04:24:39 -07:00
assert next_batch is None
assert len(generations) == 1
assert generations[0].generated_text.text == " test_get_all_users_with_"
assert generations[0].request_id == batch.requests[0].id
2023-01-20 04:24:39 -07:00
assert (
generations[0].generated_text.generated_tokens
2023-01-20 04:24:39 -07:00
== batch.stopping_criterias[0].max_new_tokens
)
2023-01-20 07:35:22 -07:00
@pytest.mark.skip
2023-01-20 04:24:39 -07:00
def test_fim_santacoder_generate_token_completion(
default_santacoder, default_fim_pb_batch
):
batch = CausalLMBatch.from_pb(
default_fim_pb_batch, default_santacoder.tokenizer, default_santacoder.device
)
next_batch = batch
for _ in range(batch.stopping_criterias[0].max_new_tokens - 1):
generations, next_batch = default_santacoder.generate_token(next_batch)
assert len(generations) == len(next_batch)
2023-01-20 04:24:39 -07:00
generations, next_batch = default_santacoder.generate_token(next_batch)
2023-01-20 04:24:39 -07:00
assert next_batch is None
assert len(generations) == 1
2023-01-20 04:24:39 -07:00
assert (
generations[0].generated_text.text
== """ineProperty(exports, "__esModule", { value"""
2023-01-20 04:24:39 -07:00
)
assert generations[0].request_id == batch.requests[0].id
2023-01-20 04:24:39 -07:00
assert (
generations[0].generated_text.generated_tokens
2023-01-20 04:24:39 -07:00
== batch.stopping_criterias[0].max_new_tokens
)