forked from Arnav0400/peft
-
Notifications
You must be signed in to change notification settings - Fork 0
/
test_decoder_models.py
118 lines (94 loc) · 5.14 KB
/
test_decoder_models.py
1
2
3
4
5
6
7
8
9
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
109
110
111
112
113
114
115
116
117
118
# coding=utf-8
# Copyright 2023-present the HuggingFace Inc. team.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import unittest
import torch
from parameterized import parameterized
from transformers import AutoModelForCausalLM
from .testing_common import PeftCommonTester, PeftTestConfigManager
PEFT_DECODER_MODELS_TO_TEST = [
"hf-internal-testing/tiny-random-OPTForCausalLM",
"hf-internal-testing/tiny-random-GPTNeoXForCausalLM",
"hf-internal-testing/tiny-random-GPT2LMHeadModel",
"hf-internal-testing/tiny-random-BloomForCausalLM",
"hf-internal-testing/tiny-random-gpt_neo",
"hf-internal-testing/tiny-random-GPTJForCausalLM",
"hf-internal-testing/tiny-random-GPTBigCodeForCausalLM",
"HuggingFaceM4/tiny-random-LlamaForCausalLM",
]
FULL_GRID = {
"model_ids": PEFT_DECODER_MODELS_TO_TEST,
"task_type": "CAUSAL_LM",
}
def skip_non_pt_mqa(test_list):
r"""
Skip tests that are prefix tuning for MQA models (not supported yet)
"""
return [test for test in test_list if not ("prefix_tuning" in test[0] and "GPTBigCodeForCausalLM" in test[0])]
class PeftDecoderModelTester(unittest.TestCase, PeftCommonTester):
r"""
Test if the PeftModel behaves as expected. This includes:
- test if the model has the expected methods
We use parametrized.expand for debugging purposes to test each model individually.
"""
transformers_class = AutoModelForCausalLM
def prepare_inputs_for_testing(self):
input_ids = torch.tensor([[1, 1, 1], [1, 2, 1]]).to(self.torch_device)
attention_mask = torch.tensor([[1, 1, 1], [1, 0, 1]]).to(self.torch_device)
input_dict = {
"input_ids": input_ids,
"attention_mask": attention_mask,
}
return input_dict
@parameterized.expand(PeftTestConfigManager.get_grid_parameters(FULL_GRID))
def test_attributes_parametrized(self, test_name, model_id, config_cls, config_kwargs):
self._test_model_attr(model_id, config_cls, config_kwargs)
@parameterized.expand(PeftTestConfigManager.get_grid_parameters(FULL_GRID))
def test_prepare_for_training_parametrized(self, test_name, model_id, config_cls, config_kwargs):
self._test_prepare_for_training(model_id, config_cls, config_kwargs)
@parameterized.expand(PeftTestConfigManager.get_grid_parameters(FULL_GRID))
def test_save_pretrained(self, test_name, model_id, config_cls, config_kwargs):
self._test_save_pretrained(model_id, config_cls, config_kwargs)
@parameterized.expand(
PeftTestConfigManager.get_grid_parameters(
{
"model_ids": PEFT_DECODER_MODELS_TO_TEST,
"lora_kwargs": {"init_lora_weights": [False]},
"task_type": "CAUSAL_LM",
},
)
)
def test_merge_layers(self, test_name, model_id, config_cls, config_kwargs):
self._test_merge_layers(model_id, config_cls, config_kwargs)
@parameterized.expand(PeftTestConfigManager.get_grid_parameters(FULL_GRID, filter_params_func=skip_non_pt_mqa))
def test_generate(self, test_name, model_id, config_cls, config_kwargs):
self._test_generate(model_id, config_cls, config_kwargs)
@parameterized.expand(PeftTestConfigManager.get_grid_parameters(FULL_GRID, filter_params_func=skip_non_pt_mqa))
def test_generate_half_prec(self, test_name, model_id, config_cls, config_kwargs):
self._test_generate_half_prec(model_id, config_cls, config_kwargs)
@parameterized.expand(PeftTestConfigManager.get_grid_parameters(FULL_GRID))
def test_training_decoders(self, test_name, model_id, config_cls, config_kwargs):
self._test_training(model_id, config_cls, config_kwargs)
@parameterized.expand(PeftTestConfigManager.get_grid_parameters(FULL_GRID))
def test_training_decoders_layer_indexing(self, test_name, model_id, config_cls, config_kwargs):
self._test_training_layer_indexing(model_id, config_cls, config_kwargs)
@parameterized.expand(PeftTestConfigManager.get_grid_parameters(FULL_GRID))
def test_training_decoders_gradient_checkpointing(self, test_name, model_id, config_cls, config_kwargs):
self._test_training_gradient_checkpointing(model_id, config_cls, config_kwargs)
@parameterized.expand(PeftTestConfigManager.get_grid_parameters(FULL_GRID))
def test_inference_safetensors(self, test_name, model_id, config_cls, config_kwargs):
self._test_inference_safetensors(model_id, config_cls, config_kwargs)
@parameterized.expand(PeftTestConfigManager.get_grid_parameters(FULL_GRID))
def test_peft_model_device_map(self, test_name, model_id, config_cls, config_kwargs):
self._test_peft_model_device_map(model_id, config_cls, config_kwargs)