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
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
|
# Copyright 2025 The HuggingFace Team. All rights reserved.
#
# 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
from unittest import skip
import torch
from torch.utils.benchmark import Timer
from accelerate.test_utils import require_huggingface_suite, require_non_cpu, require_non_hpu, slow, torch_device
from accelerate.utils import compile_regions, extract_model_from_parallel, release_memory
MODEL_ID = "gpt2"
COMPILE_ITERS = 2
INFERENCE_ITERS = 100
INFRENCE_STMT = "model(input_ids, use_cache=False)"
COMPILE_STMT = f"torch._dynamo.reset(); torch._inductor.utils.clear_inductor_caches(); {INFRENCE_STMT}"
if torch_device == "hpu":
backend = "hpu_backend"
else:
backend = "inductor"
@require_huggingface_suite
@skip("Don't work with torch 2.8")
class RegionalCompilationTester(unittest.TestCase):
def _get_model_and_inputs(self):
from transformers import AutoConfig, AutoModelForCausalLM
with torch.device(torch_device):
config = AutoConfig.from_pretrained(MODEL_ID)
model = AutoModelForCausalLM.from_config(config)
input_ids = torch.randint(0, 1000, (4, 128), dtype=torch.int64)
return model, input_ids
def test_regions_are_compiled(self):
model, _ = self._get_model_and_inputs()
compiled_model = compile_regions(model, mode="reduce-overhead", backend=backend)
# Check that the compiled model keeps a reference to the original model
assert hasattr(compiled_model, "_orig_mod")
assert compiled_model._orig_mod is model
# Check that the compiled_model.transformer.h[i] and compiled_model.lm_head are compiled separately
assert isinstance(compiled_model.transformer.h[0], torch._dynamo.eval_frame.OptimizedModule)
assert isinstance(compiled_model.lm_head, torch._dynamo.eval_frame.OptimizedModule)
assert compiled_model.transformer.h[0]._orig_mod is model.transformer.h[0]
assert compiled_model.lm_head._orig_mod is model.lm_head
def test_extract_model_keep_torch_compile(self):
model, _ = self._get_model_and_inputs()
compiled_model = compile_regions(model, mode="reduce-overhead", backend=backend)
distributed_model = torch.nn.parallel.DataParallel(model)
distributed_compiled_model = compile_regions(distributed_model, mode="reduce-overhead", backend=backend)
compiled_model_unwrapped = extract_model_from_parallel(distributed_compiled_model, keep_torch_compile=True)
assert compiled_model._orig_mod is compiled_model_unwrapped._orig_mod
def test_extract_model_remove_torch_compile(self):
model, _ = self._get_model_and_inputs()
compiled_model = compile_regions(model, mode="reduce-overhead", backend=backend)
distributed_model = torch.nn.parallel.DataParallel(model)
distributed_compiled_model = compile_regions(distributed_model, mode="reduce-overhead", backend=backend)
compiled_model_unwrapped = extract_model_from_parallel(distributed_compiled_model, keep_torch_compile=False)
assert compiled_model._orig_mod is compiled_model_unwrapped
@require_non_cpu
@require_huggingface_suite
def test_regional_compilation_cold_start(self):
model, input_ids = self._get_model_and_inputs()
regional_compilation_model = compile_regions(model, backend=backend)
regional_compilation_cold_start = (
Timer(stmt=COMPILE_STMT, globals={"model": regional_compilation_model, "input_ids": input_ids})
.timeit(COMPILE_ITERS)
.median
)
full_compilation_model = torch.compile(model, backend=backend)
full_compilation_cold_start = (
Timer(stmt=COMPILE_STMT, globals={"model": full_compilation_model, "input_ids": input_ids})
.timeit(COMPILE_ITERS)
.median
)
self.assertLess(
regional_compilation_cold_start,
full_compilation_cold_start,
"Regional compilation should have a faster cold start than full compilation",
)
release_memory(model, full_compilation_model, regional_compilation_model)
@slow
@require_non_hpu
@require_non_cpu
@require_huggingface_suite
def test_regional_compilation_inference_speedup(self):
model, input_ids = self._get_model_and_inputs()
baseline_inference_latency = (
Timer(stmt=INFRENCE_STMT, globals={"model": model, "input_ids": input_ids}).timeit(INFERENCE_ITERS).median
)
regional_compilation_model = compile_regions(model, backend=backend)
regional_compilation_inference_latency = (
Timer(stmt=INFRENCE_STMT, globals={"model": regional_compilation_model, "input_ids": input_ids})
.timeit(INFERENCE_ITERS)
.median
)
full_compilation_model = torch.compile(model, backend=backend)
full_compilation_inference_latency = (
Timer(stmt=INFRENCE_STMT, globals={"model": full_compilation_model, "input_ids": input_ids})
.timeit(INFERENCE_ITERS)
.median
)
full_compilation_inference_speedup = baseline_inference_latency / full_compilation_inference_latency
regional_compilation_inference_speedup = baseline_inference_latency / regional_compilation_inference_latency
self.assertAlmostEqual(
regional_compilation_inference_speedup,
full_compilation_inference_speedup,
delta=0.1,
msg="Regional compilation should have a similar speedup to full compilation",
)
release_memory(model, full_compilation_model, regional_compilation_model)
|
