Skip to content

GitLab

Switch branch/tag
Find file Normal viewHistoryPermalink
test_santacoder.py 3.35 KB
Newer Older
jixx's avatar
init  
jixx committed
1
2
3 
import pytest

from text_generation_server.pb import generate_pb2
jixx's avatar
add tgi2.4.0  
jixx committed
4 
from text_generation_server.models.causal_lm import CausalLMBatch, CausalLM
jixx's avatar
init  
jixx committed
5
6
7
8 


@pytest.fixture(scope="session")
def default_santacoder():
jixx's avatar
add tgi2.4.0  
jixx committed
9 
    return CausalLM.fallback(model_id="bigcode/santacoder")
jixx's avatar
init  
jixx committed
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108 


@pytest.fixture
def default_pb_request(default_pb_parameters, default_pb_stop_parameters):
    return generate_pb2.Request(
        id=0,
        inputs="def",
        input_chunks=generate_pb2.Input(chunks=[generate_pb2.InputChunk(text="def")]),
        prefill_logprobs=True,
        truncate=100,
        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>",
        input_chunks=generate_pb2.Input(
            chunks=[
                generate_pb2.InputChunk(
                    text="<fim-prefix>def<fim-suffix>world<fim-middle>"
                )
            ]
        ),
        prefill_logprobs=True,
        truncate=100,
        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)


@pytest.mark.skip
def test_santacoder_generate_token_completion(default_santacoder, default_pb_batch):
    batch = CausalLMBatch.from_pb(
        default_pb_batch,
        default_santacoder.tokenizer,
        default_santacoder.dtype,
        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)

    generations, next_batch, _ = default_santacoder.generate_token(next_batch)
    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
    assert (
        generations[0].generated_text.generated_tokens
        == batch.stopping_criterias[0].max_new_tokens
    )


@pytest.mark.skip
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.dtype,
        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)

    generations, next_batch, _ = default_santacoder.generate_token(next_batch)
    assert next_batch is None

    assert len(generations) == 1
    assert (
        generations[0].generated_text.text
        == """ineProperty(exports, "__esModule", { value"""
    )
    assert generations[0].request_id == batch.requests[0].id
    assert (
        generations[0].generated_text.generated_tokens
        == batch.stopping_criterias[0].max_new_tokens
    )