xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2025-12-10 08:34:56 +08:00
Code Issues Packages Projects Releases Wiki Activity

[Speculative Decoding] Test refactor (#8317)

Browse Source 
Co-authored-by: youkaichao <youkaichao@126.com>
This commit is contained in:
Lily Liu 2024-09-11 14:07:34 -07:00 committed by GitHub
parent 8baa454937
commit 775f00f81e
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
12 changed files with 927 additions and 1042 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
.buildkite/test-pipeline.yaml
View File

@ -217,7 +217,8 @@ steps:
commands: commands:
# See https://github.com/vllm-project/vllm/issues/5152 # See https://github.com/vllm-project/vllm/issues/5152
- export VLLM_ATTENTION_BACKEND=XFORMERS - export VLLM_ATTENTION_BACKEND=XFORMERS
- pytest -v -s spec_decode - pytest -v -s spec_decode/e2e/test_multistep_correctness.py
- pytest -v -s spec_decode --ignore=spec_decode/e2e/test_multistep_correctness.py
- label: LoRA Test %N # 30min each - label: LoRA Test %N # 30min each
mirror_hardwares: [amd] mirror_hardwares: [amd]

471
tests/spec_decode/e2e/conftest.py
View File

@ -1,224 +1,54 @@
import asyncio
import os
from itertools import cycle from itertools import cycle
from typing import Dict, List, Optional, Sequence, Tuple, Union from typing import List, Optional, Tuple
import pytest import pytest
import ray
import torch
from vllm import LLM from vllm import LLM, SamplingParams
from vllm.engine.arg_utils import AsyncEngineArgs
from vllm.engine.async_llm_engine import AsyncLLMEngine
from vllm.lora.request import LoRARequest
from vllm.model_executor.utils import set_random_seed from vllm.model_executor.utils import set_random_seed
from vllm.multimodal import MultiModalDataDict
from vllm.outputs import RequestOutput
from vllm.prompt_adapter.request import PromptAdapterRequest
from vllm.sampling_params import SamplingParams
from vllm.sequence import Logprob
from vllm.usage.usage_lib import UsageContext
from vllm.utils import Counter, random_uuid
from ...conftest import cleanup from ...conftest import cleanup
from ...utils import wait_for_gpu_memory_to_clear from ...models.utils import check_logprobs_close, check_outputs_equal
from ...utils import RemoteOpenAIServer
PROMPTS = [
class AsyncLLM: "Hello, my name is",
"""AsyncLLM "The president of the United States is",
"The capital of France is",
Note: Current LLM class in vllm don't support async mode, for test purpose, "The future of AI is",
we implement async one in here. Maybe we could move to "San Francisco is know for its",
vllm/entrypoints/llm.py in future. "Facebook was created in 2004 by",
"Curious George is a",
Below AsyncLLM is directly borrow from vllm/entrypoints/llm.py with changes "Python 3.11 brings improvements to its",
to make to work in async mode. ]
"""
def __init__(
self,
model: str,
tokenizer: Optional[str] = None,
tokenizer_mode: str = "auto",
skip_tokenizer_init: bool = False,
trust_remote_code: bool = False,
tensor_parallel_size: int = 1,
dtype: str = "auto",
quantization: Optional[str] = None,
revision: Optional[str] = None,
tokenizer_revision: Optional[str] = None,
seed: int = 0,
gpu_memory_utilization: float = 0.9,
swap_space: int = 4,
enforce_eager: bool = False,
max_seq_len_to_capture: int = 8192,
disable_custom_all_reduce: bool = False,
**kwargs,
) -> None:
if "disable_log_stats" not in kwargs:
kwargs["disable_log_stats"] = True
# Needed to engine_use_ray works as a deprecated feature,
# otherwise the following constructor will raise an exception
os.environ["VLLM_ALLOW_ENGINE_USE_RAY"] = "1"
engine_args = AsyncEngineArgs(
model=model,
tokenizer=tokenizer,
tokenizer_mode=tokenizer_mode,
skip_tokenizer_init=skip_tokenizer_init,
trust_remote_code=trust_remote_code,
tensor_parallel_size=tensor_parallel_size,
dtype=dtype,
quantization=quantization,
revision=revision,
tokenizer_revision=tokenizer_revision,
seed=seed,
gpu_memory_utilization=gpu_memory_utilization,
swap_space=swap_space,
enforce_eager=enforce_eager,
max_seq_len_to_capture=max_seq_len_to_capture,
# For now use ray for the distributed back-end, since
# we rely on the use of engine_use_ray=True to avoid
# reinitializing CUDA in the same process (driver worker)
engine_use_ray=True,
distributed_executor_backend="ray",
disable_custom_all_reduce=disable_custom_all_reduce,
**kwargs,
)
self.request_counter = Counter()
self.llm_engine = AsyncLLMEngine.from_engine_args(
engine_args, usage_context=UsageContext.LLM_CLASS)
def generate(
self,
prompts: Optional[Union[str, List[str]]] = None,
sampling_params: Optional[Union[SamplingParams,
List[SamplingParams]]] = None,
prompt_token_ids: Optional[List[List[int]]] = None,
use_tqdm: bool = True,
lora_request: Optional[LoRARequest] = None,
multi_modal_data: Optional[MultiModalDataDict] = None,
prompt_adapter_request: Optional[PromptAdapterRequest] = None
) -> List[RequestOutput]:
if prompts is None:
raise ValueError("prompts must be provided.")
if isinstance(prompts, str):
# Convert a single prompt to a list.
prompts = [prompts]
if prompts is not None:
num_requests = len(prompts)
if sampling_params is None:
# Use default sampling params.
sampling_params = SamplingParams()
elif isinstance(sampling_params,
list) and len(sampling_params) != num_requests:
raise ValueError("The lengths of prompts and "
"sampling_params must be the same.")
async def get_output(prompt, sampling_param) -> RequestOutput:
request_id = random_uuid()
results_generator = self.llm_engine.generate(
prompt, sampling_param, request_id)
final_output = None
async for request_output in results_generator:
final_output = request_output
assert final_output is not None
return final_output
outputs: List[RequestOutput] = []
try:
for i in range(num_requests):
prompt = prompts[i] if prompts is not None else None
params = sampling_params[i] if isinstance(
sampling_params, Sequence) else sampling_params
res = asyncio.run(get_output(prompt, params))
outputs.append(res)
finally:
ray.shutdown()
return outputs
@pytest.fixture @pytest.fixture
def baseline_llm_generator(request, common_llm_kwargs, def test_llm_generator(common_llm_kwargs, per_test_common_llm_kwargs,
per_test_common_llm_kwargs, baseline_llm_kwargs,
seed):
return create_llm_generator("baseline", request, common_llm_kwargs,
per_test_common_llm_kwargs,
baseline_llm_kwargs, seed)
@pytest.fixture
def test_llm_generator(request, common_llm_kwargs, per_test_common_llm_kwargs,
test_llm_kwargs, seed): test_llm_kwargs, seed):
return create_llm_generator("test", request, common_llm_kwargs,
per_test_common_llm_kwargs, test_llm_kwargs,
seed)
def generate():
kwargs = {
**common_llm_kwargs,
**per_test_common_llm_kwargs,
**test_llm_kwargs,
}
def create_llm_generator(baseline_or_test, request, common_llm_kwargs, llm = LLM(**kwargs)
per_test_common_llm_kwargs, distinct_llm_kwargs,
seed):
kwargs = {
**common_llm_kwargs,
**per_test_common_llm_kwargs,
**distinct_llm_kwargs,
}
test_name = request.node.name
model = kwargs["model"]
draft_model = kwargs.get("speculative_model", None)
same_draft_target_model = (draft_model is not None
and draft_model == model)
def generator_inner():
wait_for_gpu_memory_to_clear(
devices=list(range(torch.cuda.device_count())),
threshold_bytes=2 * 2**30,
timeout_s=60,
)
use_async = False
if "use_async" in kwargs:
use_async = kwargs.pop("use_async")
print(f'{use_async=}')
print(f'Creating {baseline_or_test=} LLM for {test_name=}. {kwargs=}')
llm = AsyncLLM(**kwargs) if use_async else LLM(**kwargs)
# Override logging interval to 0 for spec decode test run to
# log all metrics in time.
if (baseline_or_test == "test" and not use_async
and llm.llm_engine.log_stats):
for sate_logger in llm.llm_engine.stat_loggers.values():
sate_logger.local_interval = 0
if seed is not None: if seed is not None:
set_random_seed(seed) set_random_seed(seed)
yield llm yield llm
del llm del llm
cleanup() cleanup()
def generator_outer(): return generate
for llm in generator_inner():
yield llm
del llm
# Set an attribute to the generator_outer function to allow us to
# determine whether to further check the acceptance rate in tests.
generator_outer.same_draft_target_model = same_draft_target_model # type: ignore
return generator_outer
def maybe_assert_ngram_worker(llm): def maybe_assert_ngram_worker(llm):
# Verify the proposer worker is ngram if ngram is specified. # Verify the proposer worker is ngram if ngram is specified.
if (not isinstance(llm, AsyncLLM) if (llm.llm_engine.speculative_config is not None
and llm.llm_engine.speculative_config is not None
and llm.llm_engine.speculative_config.ngram_prompt_lookup_max > 0): and llm.llm_engine.speculative_config.ngram_prompt_lookup_max > 0):
from vllm.spec_decode.ngram_worker import NGramWorker from vllm.spec_decode.ngram_worker import NGramWorker
assert isinstance( assert isinstance(
@ -251,118 +81,165 @@ def get_output_from_llm_generator(
return tokens, token_ids, acceptance_rate return tokens, token_ids, acceptance_rate
def get_logprobs_from_llm_generator( def run_logprob_correctness_test(vllm_runner,
llm_generator, prompts, common_llm_kwargs,
sampling_params) -> List[List[Dict[int, Logprob]]]: per_test_common_llm_kwargs,
"""Returns a dict of (token_id: Logprob) for each generated position, for baseline_llm_kwargs,
each sequence in the batch. test_llm_kwargs,
""" batch_size: int,
for llm in llm_generator(): max_output_len: int,
outputs = llm.generate(prompts, sampling_params, use_tqdm=True) seed: Optional[int] = 0,
logprobs = [output.outputs[0].logprobs[:] for output in outputs] temperature: float = 0.0,
del llm logprobs: int = 1):
org_args = {
**common_llm_kwargs,
**per_test_common_llm_kwargs,
**baseline_llm_kwargs,
}
return logprobs sd_args = {
**common_llm_kwargs,
**per_test_common_llm_kwargs,
**test_llm_kwargs,
}
prompts = [prompt for prompt, _ in zip(cycle(PROMPTS), range(batch_size))]
sampling_params = SamplingParams(temperature=temperature,
max_tokens=max_output_len,
seed=seed,
logprobs=logprobs)
with vllm_runner(**org_args) as vllm_model:
org_outputs = vllm_model.generate_w_logprobs(prompts, sampling_params)
with vllm_runner(**sd_args) as vllm_model:
sd_outputs = vllm_model.generate_w_logprobs(prompts, sampling_params)
check_logprobs_close(outputs_0_lst=org_outputs,
outputs_1_lst=sd_outputs,
name_0="org",
name_1="sd")
def run_greedy_equality_correctness_test(baseline_llm_generator, def run_equality_correctness_test(
test_llm_generator, vllm_runner,
batch_size, common_llm_kwargs,
max_output_len, per_test_common_llm_kwargs,
force_output_len: bool, baseline_llm_kwargs,
print_tokens: bool = False, test_llm_kwargs,
ensure_all_accepted: bool = False): batch_size: int,
max_output_len: int,
seed: Optional[int] = 0,
temperature: float = 0.0,
disable_seed: bool = False,
ignore_eos: bool = True,
ensure_all_accepted: bool = False,
expected_acceptance_rate: Optional[float] = None):
org_args = {
**common_llm_kwargs,
**per_test_common_llm_kwargs,
**baseline_llm_kwargs,
}
sd_args = {
**common_llm_kwargs,
**per_test_common_llm_kwargs,
**test_llm_kwargs,
}
prompts = [prompt for prompt, _ in zip(cycle(PROMPTS), range(batch_size))]
if disable_seed:
seed = None
sampling_params = SamplingParams(temperature=temperature,
max_tokens=max_output_len,
seed=seed,
ignore_eos=ignore_eos)
with vllm_runner(**org_args) as vllm_model:
org_outputs = vllm_model.generate(prompts, sampling_params)
with vllm_runner(**sd_args) as vllm_model:
if ensure_all_accepted or expected_acceptance_rate is not None:
# Force log interval to be 0 to catch all metrics.
stat_logger = vllm_model.model.llm_engine.stat_loggers[
'prometheus']
stat_logger.local_interval = -100
sd_outputs = vllm_model.generate(prompts, sampling_params)
if ensure_all_accepted or expected_acceptance_rate is not None:
acceptance_rate = (stat_logger.metrics.
gauge_spec_decode_draft_acceptance_rate.labels(
**stat_logger.labels)._value.get())
if ensure_all_accepted:
assert True
# FIXME: ci fails to log acceptance rate.
# It works locally.
# assert acceptance_rate == 1.0
if expected_acceptance_rate is not None:
assert acceptance_rate >= expected_acceptance_rate - 1e-2
check_outputs_equal(outputs_0_lst=org_outputs,
outputs_1_lst=sd_outputs,
name_0="org",
name_1="sd")
def run_equality_correctness_test_tp(model,
common_llm_kwargs,
per_test_common_llm_kwargs,
baseline_llm_kwargs,
test_llm_kwargs,
batch_size: int,
max_output_len: int,
seed: int = 0,
temperature: float = 0.0):
"""Helper method that compares the outputs of both the baseline LLM and """Helper method that compares the outputs of both the baseline LLM and
the test LLM. It asserts greedy equality, e.g. that the outputs are exactly the test LLM. It asserts greedy equality, e.g. that the outputs are exactly
the same when temperature is zero. the same when temperature is zero.
""" """
arg1 = common_llm_kwargs + per_test_common_llm_kwargs + baseline_llm_kwargs
arg2 = common_llm_kwargs + per_test_common_llm_kwargs + test_llm_kwargs
env1 = env2 = None
run_equality_correctness_test(baseline_llm_generator, max_wait_seconds = 240
test_llm_generator, results = []
batch_size,
max_output_len,
force_output_len,
temperature=0.0,
seeded=False,
print_tokens=print_tokens,
ensure_all_accepted=ensure_all_accepted)
prompts = [prompt for prompt, _ in zip(cycle(PROMPTS), range(batch_size))]
def run_equality_correctness_test( for args, env in ((arg1, env1), (arg2, env2)):
baseline_llm_generator, with RemoteOpenAIServer(model,
test_llm_generator, args,
batch_size, env_dict=env,
max_output_len, max_wait_seconds=max_wait_seconds) as server:
force_output_len: bool, client = server.get_client()
temperature: float,
seeded: bool,
print_tokens: bool = False,
ensure_all_accepted: bool = False,
expected_acceptance_rate: Optional[float] = None):
"""Helper method that compares the outputs of both the baseline LLM and
the test LLM. It asserts greedy equality, e.g. that the outputs are exactly
the same when temperature is zero (or when temperature is > 0 and seeded).
"""
prompts = [ completion = client.completions.create(model=model,
"Hello, my name is", prompt=prompts,
"The president of the United States is", max_tokens=max_output_len,
"The capital of France is", seed=seed,
"The future of AI is", temperature=temperature)
"San Francisco is know for its",
"Facebook was created in 2004 by",
"Curious George is a",
"Python 3.11 brings improvements to its",
]
prompts = [prompt for prompt, _ in zip(cycle(prompts), range(batch_size))] results.append({
"test":
"seeded_sampling",
"text": [choice.text for choice in completion.choices],
"finish_reason":
[choice.finish_reason for choice in completion.choices],
"usage":
completion.usage,
})
# If the test requires that we generated max_output_len tokens, then set the n = len(results) // 2
# sampling params to ignore eos token. arg1_results = results[:n]
ignore_eos = force_output_len arg2_results = results[n:]
for arg1_result, arg2_result in zip(arg1_results, arg2_results):
if seeded: assert arg1_result == arg2_result, (
sampling_params = [ f"Results for {model=} are not the same with {arg1=} and {arg2=}. "
SamplingParams( f"{arg1_result=} != {arg2_result=}")
max_tokens=max_output_len,
ignore_eos=ignore_eos,
temperature=temperature,
seed=i,
) for i in range(len(prompts))
]
else:
sampling_params = SamplingParams(
max_tokens=max_output_len,
ignore_eos=ignore_eos,
temperature=temperature,
)
(spec_batch_tokens, spec_batch_token_ids,
acceptance_rate) = get_output_from_llm_generator(test_llm_generator,
prompts, sampling_params)
(baseline_batch_tokens, baseline_batch_token_ids,
_) = get_output_from_llm_generator(baseline_llm_generator, prompts,
sampling_params)
assert len(baseline_batch_token_ids) == len(prompts)
assert len(spec_batch_token_ids) == len(prompts)
for i, (baseline_token_ids, baseline_tokens, spec_token_ids,
spec_tokens) in enumerate(
zip(baseline_batch_token_ids, baseline_batch_tokens,
spec_batch_token_ids, spec_batch_tokens)):
if print_tokens:
print(f'{i=} {baseline_tokens=}')
print(f'{i=} {spec_tokens=}')
print(f'{i=} {baseline_token_ids=}')
print(f'{i=} {spec_token_ids=}')
assert baseline_token_ids == spec_token_ids
print(f'{acceptance_rate=}')
if ensure_all_accepted:
assert acceptance_rate == 1.0
if expected_acceptance_rate is not None:
assert acceptance_rate >= expected_acceptance_rate - 1e-2

97
tests/spec_decode/e2e/test_eagle_correctness.py
View File

@ -21,7 +21,7 @@ correctess for the target model outputs.
import pytest import pytest
from .conftest import run_greedy_equality_correctness_test from .conftest import run_equality_correctness_test
# main model # main model
MAIN_MODEL = "JackFram/llama-68m" MAIN_MODEL = "JackFram/llama-68m"
@ -53,7 +53,7 @@ PRECISION = "float32"
"dtype": PRECISION, "dtype": PRECISION,
# Main model # Main model
"model": MAIN_MODEL, "model_name": MAIN_MODEL,
}]) }])
@pytest.mark.parametrize("per_test_common_llm_kwargs", [{}]) @pytest.mark.parametrize("per_test_common_llm_kwargs", [{}])
@pytest.mark.parametrize("baseline_llm_kwargs", [{}]) @pytest.mark.parametrize("baseline_llm_kwargs", [{}])
@ -68,15 +68,16 @@ PRECISION = "float32"
]) ])
@pytest.mark.parametrize("batch_size", [1, 32]) @pytest.mark.parametrize("batch_size", [1, 32])
@pytest.mark.parametrize("seed", [1]) @pytest.mark.parametrize("seed", [1])
def test_eagle_e2e_greedy_correctness(baseline_llm_generator, def test_eagle_e2e_greedy_correctness(vllm_runner, common_llm_kwargs,
test_llm_generator, batch_size: int, per_test_common_llm_kwargs,
output_len: int): baseline_llm_kwargs, test_llm_kwargs,
"""Verify greedy equality with different batch size.""" batch_size: int, output_len: int,
run_greedy_equality_correctness_test(baseline_llm_generator, seed: int):
test_llm_generator,
batch_size, run_equality_correctness_test(vllm_runner, common_llm_kwargs,
max_output_len=output_len, per_test_common_llm_kwargs,
force_output_len=True) baseline_llm_kwargs, test_llm_kwargs,
batch_size, output_len, seed)
@pytest.mark.parametrize( @pytest.mark.parametrize(
@ -94,7 +95,7 @@ def test_eagle_e2e_greedy_correctness(baseline_llm_generator,
"dtype": PRECISION, "dtype": PRECISION,
# Main model # Main model
"model": MAIN_MODEL, "model_name": MAIN_MODEL,
}]) }])
@pytest.mark.parametrize("per_test_common_llm_kwargs", [{}]) @pytest.mark.parametrize("per_test_common_llm_kwargs", [{}])
@pytest.mark.parametrize("baseline_llm_kwargs", [{}]) @pytest.mark.parametrize("baseline_llm_kwargs", [{}])
@ -109,17 +110,16 @@ def test_eagle_e2e_greedy_correctness(baseline_llm_generator,
]) ])
@pytest.mark.parametrize("batch_size", [1, 32]) @pytest.mark.parametrize("batch_size", [1, 32])
@pytest.mark.parametrize("seed", [1]) @pytest.mark.parametrize("seed", [1])
def test_eagle_e2e_greedy_correctness_cuda_graph(baseline_llm_generator, def test_eagle_e2e_greedy_correctness_cuda_graph(
test_llm_generator, vllm_runner, common_llm_kwargs, per_test_common_llm_kwargs,
batch_size: int, baseline_llm_kwargs, test_llm_kwargs, batch_size: int, output_len: int,
output_len: int): seed: int):
"""Verify greedy equality with cuda graph enabled and different """Verify greedy equality with cuda graph enabled and different
batch sizes.""" batch sizes."""
run_greedy_equality_correctness_test(baseline_llm_generator, run_equality_correctness_test(vllm_runner, common_llm_kwargs,
test_llm_generator, per_test_common_llm_kwargs,
batch_size, baseline_llm_kwargs, test_llm_kwargs,
max_output_len=output_len, batch_size, output_len, seed)
force_output_len=True)
@pytest.mark.parametrize( @pytest.mark.parametrize(
@ -140,7 +140,7 @@ def test_eagle_e2e_greedy_correctness_cuda_graph(baseline_llm_generator,
"dtype": PRECISION, "dtype": PRECISION,
# Main model # Main model
"model": MAIN_MODEL, "model_name": MAIN_MODEL,
}]) }])
@pytest.mark.parametrize("per_test_common_llm_kwargs", [{}]) @pytest.mark.parametrize("per_test_common_llm_kwargs", [{}])
@pytest.mark.parametrize("baseline_llm_kwargs", [{}]) @pytest.mark.parametrize("baseline_llm_kwargs", [{}])
@ -158,18 +158,17 @@ def test_eagle_e2e_greedy_correctness_cuda_graph(baseline_llm_generator,
]) ])
@pytest.mark.parametrize("batch_size", [4]) @pytest.mark.parametrize("batch_size", [4])
@pytest.mark.parametrize("seed", [1]) @pytest.mark.parametrize("seed", [1])
def test_eagle_e2e_greedy_correctness_with_preemption(baseline_llm_generator, def test_eagle_e2e_greedy_correctness_with_preemption(
test_llm_generator, vllm_runner, common_llm_kwargs, per_test_common_llm_kwargs,
batch_size: int, baseline_llm_kwargs, test_llm_kwargs, batch_size: int, output_len: int,
output_len: int): seed: int):
"""Verify greedy equality, even when some sequences are preempted mid- """Verify greedy equality, even when some sequences are preempted mid-
generation. generation.
""" """
run_greedy_equality_correctness_test(baseline_llm_generator, run_equality_correctness_test(vllm_runner, common_llm_kwargs,
test_llm_generator, per_test_common_llm_kwargs,
batch_size, baseline_llm_kwargs, test_llm_kwargs,
max_output_len=output_len, batch_size, output_len, seed)
force_output_len=True)
@pytest.mark.parametrize( @pytest.mark.parametrize(
@ -185,7 +184,7 @@ def test_eagle_e2e_greedy_correctness_with_preemption(baseline_llm_generator,
"dtype": PRECISION, "dtype": PRECISION,
# Main model # Main model
"model": MAIN_MODEL, "model_name": MAIN_MODEL,
}]) }])
@pytest.mark.parametrize("per_test_common_llm_kwargs", [{}]) @pytest.mark.parametrize("per_test_common_llm_kwargs", [{}])
@pytest.mark.parametrize("baseline_llm_kwargs", [{}]) @pytest.mark.parametrize("baseline_llm_kwargs", [{}])
@ -207,16 +206,17 @@ def test_eagle_e2e_greedy_correctness_with_preemption(baseline_llm_generator,
32, 32,
]) ])
@pytest.mark.parametrize("seed", [1]) @pytest.mark.parametrize("seed", [1])
def test_eagle_different_k(baseline_llm_generator, test_llm_generator, def test_eagle_different_k(vllm_runner, common_llm_kwargs,
batch_size: int, output_len: int): per_test_common_llm_kwargs, baseline_llm_kwargs,
test_llm_kwargs, batch_size: int, output_len: int,
seed: int):
"""Verify that eagle speculative decoding produces exact equality """Verify that eagle speculative decoding produces exact equality
to without spec decode with different values of num_speculative_tokens. to without spec decode with different values of num_speculative_tokens.
""" """
run_greedy_equality_correctness_test(baseline_llm_generator, run_equality_correctness_test(vllm_runner, common_llm_kwargs,
test_llm_generator, per_test_common_llm_kwargs,
batch_size, baseline_llm_kwargs, test_llm_kwargs,
max_output_len=output_len, batch_size, output_len, seed)
force_output_len=True)
@pytest.mark.parametrize( @pytest.mark.parametrize(
@ -232,7 +232,7 @@ def test_eagle_different_k(baseline_llm_generator, test_llm_generator,
"dtype": PRECISION, "dtype": PRECISION,
# Main model # Main model
"model": MAIN_MODEL, "model_name": MAIN_MODEL,
}]) }])
@pytest.mark.parametrize("per_test_common_llm_kwargs", [{}]) @pytest.mark.parametrize("per_test_common_llm_kwargs", [{}])
@pytest.mark.parametrize("baseline_llm_kwargs", [{}]) @pytest.mark.parametrize("baseline_llm_kwargs", [{}])
@ -250,17 +250,18 @@ def test_eagle_different_k(baseline_llm_generator, test_llm_generator,
32, 32,
]) ])
@pytest.mark.parametrize("seed", [1]) @pytest.mark.parametrize("seed", [1])
def test_eagle_disable_queue(baseline_llm_generator, test_llm_generator, def test_eagle_disable_queue(vllm_runner, common_llm_kwargs,
batch_size: int, output_len: int): per_test_common_llm_kwargs, baseline_llm_kwargs,
test_llm_kwargs, batch_size: int, output_len: int,
seed: int):
"""Verify that eagle speculative decoding produces exact equality """Verify that eagle speculative decoding produces exact equality
to without spec decode when speculation is disabled for large to without spec decode when speculation is disabled for large
batch sizes. batch sizes.
""" """
run_greedy_equality_correctness_test(baseline_llm_generator, run_equality_correctness_test(vllm_runner, common_llm_kwargs,
test_llm_generator, per_test_common_llm_kwargs,
batch_size, baseline_llm_kwargs, test_llm_kwargs,
max_output_len=output_len, batch_size, output_len, seed)
force_output_len=True)
if __name__ == "__main__": if __name__ == "__main__":

52
tests/spec_decode/e2e/test_integration.py
View File

@ -4,7 +4,9 @@ other features, e.g. cuda graphs.
import pytest import pytest
from .conftest import run_greedy_equality_correctness_test from .conftest import run_equality_correctness_test
MAIN_MODEL = "JackFram/llama-68m"
@pytest.mark.parametrize( @pytest.mark.parametrize(
@ -15,7 +17,7 @@ from .conftest import run_greedy_equality_correctness_test
# Verify equality when cuda graphs allowed. # Verify equality when cuda graphs allowed.
"enforce_eager": False, "enforce_eager": False,
"model": "JackFram/llama-68m", "model_name": "JackFram/llama-68m",
}]) }])
@pytest.mark.parametrize( @pytest.mark.parametrize(
"per_test_common_llm_kwargs", "per_test_common_llm_kwargs",
@ -31,23 +33,27 @@ from .conftest import run_greedy_equality_correctness_test
@pytest.mark.parametrize("batch_size", [8]) @pytest.mark.parametrize("batch_size", [8])
@pytest.mark.parametrize("output_len", [32]) @pytest.mark.parametrize("output_len", [32])
@pytest.mark.parametrize("seed", [1]) @pytest.mark.parametrize("seed", [1])
def test_spec_decode_cuda_graph(baseline_llm_generator, test_llm_generator, def test_spec_decode_cuda_graph(vllm_runner, common_llm_kwargs,
batch_size, output_len): per_test_common_llm_kwargs,
baseline_llm_kwargs, test_llm_kwargs,
batch_size: int, output_len: int, seed: int):
"""Verify spec decode equality when cuda graphs are enabled. """Verify spec decode equality when cuda graphs are enabled.
""" """
run_greedy_equality_correctness_test( run_equality_correctness_test(vllm_runner,
baseline_llm_generator, common_llm_kwargs,
test_llm_generator, per_test_common_llm_kwargs,
batch_size, baseline_llm_kwargs,
max_output_len=output_len, test_llm_kwargs,
force_output_len=True, batch_size,
) max_output_len=output_len,
seed=seed,
temperature=0.0)
@pytest.mark.parametrize( @pytest.mark.parametrize(
"common_llm_kwargs", "common_llm_kwargs",
[{ [{
"model": "JackFram/llama-160m", "model_name": "JackFram/llama-160m",
# Skip cuda graph recording for fast test. # Skip cuda graph recording for fast test.
"enforce_eager": True, "enforce_eager": True,
@ -80,13 +86,19 @@ def test_spec_decode_cuda_graph(baseline_llm_generator, test_llm_generator,
@pytest.mark.parametrize("baseline_llm_kwargs", [{}]) @pytest.mark.parametrize("baseline_llm_kwargs", [{}])
@pytest.mark.parametrize("batch_size", [2]) @pytest.mark.parametrize("batch_size", [2])
@pytest.mark.parametrize("seed", [1]) @pytest.mark.parametrize("seed", [1])
def test_speculative_model_quantization_config(baseline_llm_generator, def test_speculative_model_quantization_config(vllm_runner, common_llm_kwargs,
test_llm_generator, per_test_common_llm_kwargs,
batch_size: int): baseline_llm_kwargs,
test_llm_kwargs,
batch_size: int, seed: int):
"""Verify spec decode works well with draft model quantization configs. """Verify spec decode works well with draft model quantization configs.
""" """
run_greedy_equality_correctness_test(baseline_llm_generator, run_equality_correctness_test(vllm_runner,
test_llm_generator, common_llm_kwargs,
batch_size, per_test_common_llm_kwargs,
max_output_len=32, baseline_llm_kwargs,
force_output_len=True) test_llm_kwargs,
batch_size,
max_output_len=32,
seed=seed,
temperature=0.0)

155
tests/spec_decode/e2e/test_integration_dist_tp2.py
View File

@ -7,42 +7,39 @@ import torch
from vllm.utils import is_hip from vllm.utils import is_hip
from .conftest import run_greedy_equality_correctness_test from .conftest import run_equality_correctness_test_tp
@pytest.mark.skipif(torch.cuda.device_count() < 2, @pytest.mark.skipif(torch.cuda.device_count() < 2,
reason="Need at least 2 GPUs to run the test.") reason="Need at least 2 GPUs to run the test.")
@pytest.mark.parametrize( @pytest.mark.parametrize(
"common_llm_kwargs", "common_llm_kwargs",
[{ [[
"model": "JackFram/llama-68m",
# Skip cuda graph recording for fast test. # Skip cuda graph recording for fast test.
"enforce_eager": True, "--enforce-eager",
# Required for spec decode. # Required for spec decode.
"use_v2_block_manager": True, "--use-v2-block-manager",
"tensor_parallel_size": 2, "--tensor-parallel-size",
"2"
# Use AsyncLLM engine, so that the engine runs in its own process. ]])
# Otherwise, since vLLM does not follow true SPMD, the test runner @pytest.mark.parametrize("per_test_common_llm_kwargs", [[]])
# process will have both the engine and the rank0 worker. NCCL is not @pytest.mark.parametrize("baseline_llm_kwargs", [[]])
# cleaned up properly, and its server host thread leaks, causing the
# second run of the test to fail with internal NCCL error.
"use_async": True,
}])
@pytest.mark.parametrize("per_test_common_llm_kwargs", [{}])
@pytest.mark.parametrize("baseline_llm_kwargs", [{}])
@pytest.mark.parametrize("test_llm_kwargs", [ @pytest.mark.parametrize("test_llm_kwargs", [
{ [
"speculative_model": "JackFram/llama-68m", "--speculative-model",
"num_speculative_tokens": 3, "JackFram/llama-68m",
}, "--num-speculative-tokens",
{ "3",
"speculative_model": "[ngram]", ],
"num_speculative_tokens": 5, [
"ngram_prompt_lookup_max": 3, "--speculative-model",
}, "[ngram]",
"--num-speculative-tokens",
"5",
"--ngram-prompt-lookup-max",
"3",
],
]) ])
@pytest.mark.parametrize("batch_size", [2]) @pytest.mark.parametrize("batch_size", [2])
@pytest.mark.parametrize( @pytest.mark.parametrize(
@ -52,75 +49,75 @@ from .conftest import run_greedy_equality_correctness_test
32, 32,
]) ])
@pytest.mark.parametrize("seed", [1]) @pytest.mark.parametrize("seed", [1])
def test_target_model_tp_gt_1(baseline_llm_generator, test_llm_generator, def test_target_model_tp_gt_1(common_llm_kwargs, per_test_common_llm_kwargs,
batch_size: int, output_len: int): baseline_llm_kwargs, test_llm_kwargs,
batch_size: int, output_len: int, seed: int):
"""Verify greedy equality when tensor parallelism is used. """Verify greedy equality when tensor parallelism is used.
""" """
if is_hip(): if is_hip():
pytest.skip("hip is not well-supported yet") pytest.skip("hip is not well-supported yet")
run_greedy_equality_correctness_test(baseline_llm_generator, run_equality_correctness_test_tp("JackFram/llama-68m",
test_llm_generator, common_llm_kwargs,
batch_size, per_test_common_llm_kwargs,
max_output_len=output_len, baseline_llm_kwargs,
force_output_len=True) test_llm_kwargs,
batch_size,
output_len,
seed,
temperature=0.0)
@pytest.mark.skipif(torch.cuda.device_count() < 2, @pytest.mark.skipif(torch.cuda.device_count() < 2,
reason="Need at least 2 GPUs to run the test.") reason="Need at least 2 GPUs to run the test.")
@pytest.mark.parametrize( @pytest.mark.parametrize(
"common_llm_kwargs", "common_llm_kwargs",
[{ [[
# Skip cuda graph recording for fast test. # Skip cuda graph recording for fast test.
"enforce_eager": True, "--enforce-eager",
# Required for spec decode. # Required for spec decode.
"use_v2_block_manager": True, "--use_v2_block_manager",
"tensor_parallel_size": 2, "--tensor_parallel_size",
"2",
# Use AsyncLLM engine, so that the engine runs in its own process.
# Otherwise, since vLLM does not follow true SPMD, the test runner
# process will have both the engine and the rank0 worker. NCCL is not
# cleaned up properly, and its server host thread leaks, causing the
# second run of the test to fail with internal NCCL error.
"use_async": True,
# precision # precision
"dtype": "float32", "--dtype",
}]) "bfloat16",
@pytest.mark.parametrize("baseline_llm_kwargs", [{}]) ]])
@pytest.mark.parametrize( @pytest.mark.parametrize("per_test_common_llm_kwargs", [[]])
"per_test_common_llm_kwargs, test_llm_kwargs", @pytest.mark.parametrize("baseline_llm_kwargs", [[]])
[ @pytest.mark.parametrize("model, test_llm_kwargs",
( [("JackFram/llama-68m", [
{ "--speculative-model",
# Use a small model for a fast test. "JackFram/llama-68m",
# Note this is repeated in the test body; to initialize a "--num_speculative-tokens",
# tokenizer. "5",
"model": "JackFram/llama-68m", "--speculative-draft-tensor-parallel-size",
}, "1",
{ ]),
"speculative_model": "JackFram/llama-68m", ("ibm-granite/granite-3b-code-instruct", [
"num_speculative_tokens": 5, "--speculative-model",
"speculative_draft_tensor_parallel_size": 1, "ibm-granite/granite-3b-code-instruct",
}), "--num_speculative-tokens",
({ "5",
"model": "ibm-granite/granite-3b-code-instruct", "--speculative-draft-tensor-parallel-size",
}, { "1",
"speculative_model": ])])
"ibm-granite/granite-3b-code-instruct-accelerator",
"num_speculative_tokens": 5,
"speculative_draft_tensor_parallel_size": 1,
})
])
@pytest.mark.parametrize("batch_size", [2]) @pytest.mark.parametrize("batch_size", [2])
@pytest.mark.parametrize("seed", [1]) @pytest.mark.parametrize("seed", [1])
def test_draft_model_tp_lt_target_model_tp2(test_llm_generator, def test_draft_model_tp_lt_target_model_tp2(model, common_llm_kwargs,
baseline_llm_generator, per_test_common_llm_kwargs,
batch_size: int): baseline_llm_kwargs,
test_llm_kwargs, batch_size: int,
seed: int):
"""Verify spec decode works well with smaller tp for draft models. """Verify spec decode works well with smaller tp for draft models.
""" """
run_greedy_equality_correctness_test(baseline_llm_generator, run_equality_correctness_test_tp(model,
test_llm_generator, common_llm_kwargs,
batch_size, per_test_common_llm_kwargs,
max_output_len=32, baseline_llm_kwargs,
force_output_len=True) test_llm_kwargs,
batch_size,
max_output_len=32,
seed=seed,
temperature=0.0)

126
tests/spec_decode/e2e/test_integration_dist_tp4.py
View File

@ -2,98 +2,97 @@
tensor parallelism. tensor parallelism.
""" """
import openai
import pytest import pytest
import torch import torch
from .conftest import run_greedy_equality_correctness_test from .conftest import run_equality_correctness_test_tp
MAIN_MODEL = "JackFram/llama-68m"
SPEC_MODEL = "JackFram/llama-68m"
@pytest.mark.skipif(torch.cuda.device_count() < 4, @pytest.mark.skipif(torch.cuda.device_count() < 4,
reason="Need at least 4 GPUs to run the test.") reason="Need at least 4 GPUs to run the test.")
@pytest.mark.parametrize( @pytest.mark.parametrize(
"common_llm_kwargs", "common_llm_kwargs",
[{ [[
# Use a small model for a fast test.
# Note this is repeated in the test body; to initialize a tokenizer.
"model": "JackFram/llama-68m",
# Skip cuda graph recording for fast test. # Skip cuda graph recording for fast test.
"enforce_eager": True, "--enforce_eager",
# Required for spec decode. # Required for spec decode.
"use_v2_block_manager": True, "--use-v2-block-manager",
"tensor_parallel_size": 4, "--tensor-parallel-size",
"4",
# Use AsyncLLM engine, so that the engine runs in its own process. ]])
# Otherwise, since vLLM does not follow true SPMD, the test runner
# process will have both the engine and the rank0 worker. NCCL is not
# cleaned up properly, and its server host thread leaks, causing the
# second run of the test to fail with internal NCCL error.
"use_async": True,
}])
@pytest.mark.parametrize("per_test_common_llm_kwargs", [ @pytest.mark.parametrize("per_test_common_llm_kwargs", [
{ [
"speculative_model": "JackFram/llama-68m", "--speculative-model",
"num_speculative_tokens": 5, f"{SPEC_MODEL}",
}, "--num-speculative-tokens",
"5",
],
]) ])
@pytest.mark.parametrize("baseline_llm_kwargs", [{}]) @pytest.mark.parametrize("baseline_llm_kwargs", [[]])
@pytest.mark.parametrize( @pytest.mark.parametrize(
"test_llm_kwargs", "test_llm_kwargs",
[ [
#TODO(wooyeon): add spec_draft_dp=2 case #TODO(wooyeon): add spec_draft_dp=2 case
{ [
"speculative_draft_tensor_parallel_size": 1, "--speculative-draft-tensor-parallel-size",
}, "1",
],
]) ])
@pytest.mark.parametrize("batch_size", [2]) @pytest.mark.parametrize("batch_size", [2])
@pytest.mark.parametrize("seed", [1]) @pytest.mark.parametrize("seed", [1])
def test_draft_model_tp_lt_target_model_tp4(test_llm_generator, def test_draft_model_tp_lt_target_model_tp4(common_llm_kwargs,
baseline_llm_generator, per_test_common_llm_kwargs,
batch_size: int): baseline_llm_kwargs,
test_llm_kwargs, batch_size: int,
seed: int):
"""Verify spec decode works well with smaller tp for draft models. """Verify spec decode works well with smaller tp for draft models.
""" """
run_greedy_equality_correctness_test(baseline_llm_generator, run_equality_correctness_test_tp(MAIN_MODEL,
test_llm_generator, common_llm_kwargs,
batch_size, per_test_common_llm_kwargs,
max_output_len=32, baseline_llm_kwargs,
force_output_len=True) test_llm_kwargs,
batch_size,
max_output_len=32,
seed=seed,
temperature=0.0)
@pytest.mark.skipif(torch.cuda.device_count() < 4, @pytest.mark.skipif(torch.cuda.device_count() < 4,
reason="Need at least 4 GPUs to run the test.") reason="Need at least 4 GPUs to run the test.")
@pytest.mark.parametrize( @pytest.mark.parametrize(
"common_llm_kwargs", "common_llm_kwargs",
[{ [[
"model": "JackFram/llama-160m",
# Skip cuda graph recording for fast test. # Skip cuda graph recording for fast test.
"enforce_eager": True, "--enforce-eager",
# Required for spec decode. # Required for spec decode.
"use_v2_block_manager": True, "--use-v2-block-manager",
"tensor_parallel_size": 4, "--tensor-parallel-size",
"4",
# Use AsyncLLM engine, so that the engine runs in its own process. ]])
# Otherwise, since vLLM does not follow true SPMD, the test runner @pytest.mark.parametrize("per_test_common_llm_kwargs", [[]])
# process will have both the engine and the rank0 worker. NCCL is not @pytest.mark.parametrize("baseline_llm_kwargs", [[]])
# cleaned up properly, and its server host thread leaks, causing the
# second run of the test to fail with internal NCCL error.
"use_async": True,
}])
@pytest.mark.parametrize("per_test_common_llm_kwargs", [{}])
@pytest.mark.parametrize("baseline_llm_kwargs", [{}])
@pytest.mark.parametrize( @pytest.mark.parametrize(
"test_llm_kwargs", "test_llm_kwargs",
[ [
{ [
"speculative_model": "JackFram/llama-68m", "--speculative-model",
"num_speculative_tokens": 5, f"{SPEC_MODEL}",
"--num-speculative-tokens",
"5",
# Artificially limit the draft model max model len; this forces vLLM # Artificially limit the draft model max model len; this forces vLLM
# to skip speculation once the sequences grow beyond 32-k tokens. # to skip speculation once the sequences grow beyond 32-k tokens.
"speculative_max_model_len": 32, "--speculative-max-model-len",
}, "32",
],
]) ])
@pytest.mark.parametrize("batch_size", [8]) @pytest.mark.parametrize("batch_size", [8])
@pytest.mark.parametrize( @pytest.mark.parametrize(
@ -105,8 +104,9 @@ def test_draft_model_tp_lt_target_model_tp4(test_llm_generator,
64, 64,
]) ])
@pytest.mark.parametrize("seed", [1]) @pytest.mark.parametrize("seed", [1])
def test_skip_speculation(baseline_llm_generator, test_llm_generator, def test_skip_speculation(common_llm_kwargs, per_test_common_llm_kwargs,
batch_size: int, output_len: int): baseline_llm_kwargs, test_llm_kwargs,
batch_size: int, output_len: int, seed: int):
"""Verify job failure with RuntimeError when all sequences skip speculation. """Verify job failure with RuntimeError when all sequences skip speculation.
We do this by setting the max model len of the draft model to an We do this by setting the max model len of the draft model to an
artificially low value, such that when the sequences grow beyond it, they artificially low value, such that when the sequences grow beyond it, they
@ -114,9 +114,13 @@ def test_skip_speculation(baseline_llm_generator, test_llm_generator,
TODO: fix it to pass without raising Error. (#5814) TODO: fix it to pass without raising Error. (#5814)
""" """
with pytest.raises(RuntimeError): with pytest.raises(openai.APIConnectionError):
run_greedy_equality_correctness_test(baseline_llm_generator, run_equality_correctness_test_tp(MAIN_MODEL,
test_llm_generator, common_llm_kwargs,
batch_size, per_test_common_llm_kwargs,
max_output_len=output_len, baseline_llm_kwargs,
force_output_len=True) test_llm_kwargs,
batch_size,
output_len,
seed,
temperature=0.0)

327
tests/spec_decode/e2e/test_logprobs.py
View File

@ -1,24 +1,22 @@
import math
from itertools import cycle from itertools import cycle
import pytest import pytest
from vllm import SamplingParams from vllm import SamplingParams
from .conftest import get_logprobs_from_llm_generator from .conftest import run_logprob_correctness_test
@pytest.mark.parametrize( @pytest.mark.parametrize(
"common_llm_kwargs", "common_llm_kwargs",
[{ [{
"model": "JackFram/llama-68m", "model_name": "JackFram/llama-68m",
# Skip cuda graph recording for fast test. # Skip cuda graph recording for fast test.
"enforce_eager": True, "enforce_eager": True,
# Required for spec decode. # Required for spec decode.
"use_v2_block_manager": True, "use_v2_block_manager": True,
"max_logprobs": 6,
}]) }])
@pytest.mark.parametrize("per_test_common_llm_kwargs", [{}]) @pytest.mark.parametrize("per_test_common_llm_kwargs", [{}])
@pytest.mark.parametrize("baseline_llm_kwargs", [{}]) @pytest.mark.parametrize("baseline_llm_kwargs", [{}])
@ -36,64 +34,29 @@ from .conftest import get_logprobs_from_llm_generator
7, 7,
]) ])
@pytest.mark.parametrize("seed", [1]) @pytest.mark.parametrize("seed", [1])
def test_logprobs_equality(baseline_llm_generator, test_llm_generator, @pytest.mark.parametrize("logprobs", [1, 6])
batch_size: int, output_len: int): def test_logprobs_equality(vllm_runner, common_llm_kwargs,
per_test_common_llm_kwargs, baseline_llm_kwargs,
test_llm_kwargs, batch_size: int, output_len: int,
seed: int, logprobs: int):
"""Verify output logprobs are equal with and without speculative decoding. """Verify output logprobs are equal with and without speculative decoding.
""" """
run_greedy_logprobs_correctness_test(baseline_llm_generator, run_logprob_correctness_test(vllm_runner,
test_llm_generator, common_llm_kwargs,
batch_size, per_test_common_llm_kwargs,
max_output_len=output_len, baseline_llm_kwargs,
force_output_len=True) test_llm_kwargs,
batch_size,
output_len,
seed,
temperature=0.0,
logprobs=logprobs)
@pytest.mark.parametrize( @pytest.mark.parametrize(
"common_llm_kwargs", "common_llm_kwargs",
[{ [{
"model": "JackFram/llama-68m", "model_name": "JackFram/llama-68m",
# Skip cuda graph recording for fast test.
"enforce_eager": True,
# Required for spec decode.
"use_v2_block_manager": True,
"max_logprobs": 6,
}])
@pytest.mark.parametrize("per_test_common_llm_kwargs", [{}])
@pytest.mark.parametrize("baseline_llm_kwargs", [{}])
@pytest.mark.parametrize("test_llm_kwargs",
[{
"speculative_model": "JackFram/llama-160m",
"num_speculative_tokens": 3,
"disable_logprobs_during_spec_decoding": False,
}])
@pytest.mark.parametrize("batch_size", [1])
@pytest.mark.parametrize("num_logprobs", [6])
@pytest.mark.parametrize(
"output_len",
[
# Use smaller output len for fast test.
7,
])
@pytest.mark.parametrize("seed", [1])
def test_diff_num_logprobs(baseline_llm_generator, test_llm_generator,
batch_size: int, output_len: int,
num_logprobs: int):
"""Verify output logprobs are equal with and without spec decode.
This specifies a number of logprobs >1.
"""
run_greedy_logprobs_correctness_test(baseline_llm_generator,
test_llm_generator,
batch_size,
max_output_len=output_len,
force_output_len=True,
logprob_rank=num_logprobs)
@pytest.mark.parametrize(
"common_llm_kwargs",
[{
"model": "JackFram/llama-68m",
# Skip cuda graph recording for fast test. # Skip cuda graph recording for fast test.
"enforce_eager": True, "enforce_eager": True,
@ -121,21 +84,29 @@ def test_diff_num_logprobs(baseline_llm_generator, test_llm_generator,
32, 32,
]) ])
@pytest.mark.parametrize("seed", [1]) @pytest.mark.parametrize("seed", [1])
def test_logprobs_different_k(baseline_llm_generator, test_llm_generator, @pytest.mark.parametrize("logprobs", [1, 6])
batch_size: int, output_len: int): def test_logprobs_different_k(vllm_runner, common_llm_kwargs,
per_test_common_llm_kwargs, baseline_llm_kwargs,
test_llm_kwargs, batch_size: int,
output_len: int, seed: int, logprobs: int):
"""Veriy logprob greedy equality with different speculation lens. """Veriy logprob greedy equality with different speculation lens.
""" """
run_greedy_logprobs_correctness_test(baseline_llm_generator, run_logprob_correctness_test(vllm_runner,
test_llm_generator, common_llm_kwargs,
batch_size, per_test_common_llm_kwargs,
max_output_len=output_len, baseline_llm_kwargs,
force_output_len=True) test_llm_kwargs,
batch_size,
output_len,
seed,
temperature=0.0,
logprobs=logprobs)
@pytest.mark.parametrize( @pytest.mark.parametrize(
"common_llm_kwargs", "common_llm_kwargs",
[{ [{
"model": "JackFram/llama-68m", "model_name": "JackFram/llama-68m",
# Skip cuda graph recording for fast test. # Skip cuda graph recording for fast test.
"enforce_eager": True, "enforce_eager": True,
@ -164,22 +135,30 @@ def test_logprobs_different_k(baseline_llm_generator, test_llm_generator,
32, 32,
]) ])
@pytest.mark.parametrize("seed", [1]) @pytest.mark.parametrize("seed", [1])
def test_logprobs_when_skip_speculation(baseline_llm_generator, @pytest.mark.parametrize("logprobs", [1])
test_llm_generator, batch_size: int, def test_logprobs_when_skip_speculation(vllm_runner, common_llm_kwargs,
output_len: int): per_test_common_llm_kwargs,
baseline_llm_kwargs, test_llm_kwargs,
batch_size: int, output_len: int,
seed: int, logprobs: int):
"""Verify logprobs greedy equality when some sequences skip speculation. """Verify logprobs greedy equality when some sequences skip speculation.
""" """
run_greedy_logprobs_correctness_test(baseline_llm_generator, run_logprob_correctness_test(vllm_runner,
test_llm_generator, common_llm_kwargs,
batch_size, per_test_common_llm_kwargs,
max_output_len=output_len, baseline_llm_kwargs,
force_output_len=True) test_llm_kwargs,
batch_size,
output_len,
seed,
temperature=0.0,
logprobs=logprobs)
@pytest.mark.parametrize( @pytest.mark.parametrize(
"common_llm_kwargs", "common_llm_kwargs",
[{ [{
"model": "JackFram/llama-68m", "model_name": "JackFram/llama-68m",
# Skip cuda graph recording for fast test. # Skip cuda graph recording for fast test.
"enforce_eager": True, "enforce_eager": True,
@ -203,19 +182,17 @@ def test_logprobs_when_skip_speculation(baseline_llm_generator,
32, 32,
]) ])
@pytest.mark.parametrize("seed", [1]) @pytest.mark.parametrize("seed", [1])
def test_logprobs_temp_1(baseline_llm_generator, test_llm_generator, @pytest.mark.parametrize("logprobs", [6])
batch_size: int, output_len: int): def test_logprobs_temp_1(vllm_runner, common_llm_kwargs,
per_test_common_llm_kwargs, baseline_llm_kwargs,
test_llm_kwargs, batch_size: int, output_len: int,
seed: int, logprobs: int):
"""Verify at least one logprob result has num_logprobs+1, which tests the """Verify at least one logprob result has num_logprobs+1, which tests the
case where the sampled token is not in top-k logprobs. case where the sampled token is not in top-k logprobs.
Ideally, this test should validate equality with non-spec by getting Ideally, this test should validate equality with non-spec by getting
logprobs. This is left as future improvement. logprobs. This is left as future improvement.
""" """
batch_size = 8
max_output_len = output_len
force_output_len = True
logprob_rank = 5
temperature = 1.0 temperature = 1.0
prompts = [ prompts = [
@ -231,129 +208,40 @@ def test_logprobs_temp_1(baseline_llm_generator, test_llm_generator,
prompts = [prompt for prompt, _ in zip(cycle(prompts), range(batch_size))] prompts = [prompt for prompt, _ in zip(cycle(prompts), range(batch_size))]
# If the test requires that we generated max_output_len tokens, then set the
# sampling params to ignore eos token.
ignore_eos = force_output_len
sampling_params = SamplingParams( sampling_params = SamplingParams(
max_tokens=max_output_len, max_tokens=output_len,
ignore_eos=ignore_eos, ignore_eos=True,
temperature=temperature, temperature=temperature,
logprobs=logprob_rank, logprobs=logprobs,
) )
spec_batch_logprobs = get_logprobs_from_llm_generator( sd_args = {
test_llm_generator, prompts, sampling_params) **common_llm_kwargs,
**per_test_common_llm_kwargs,
**test_llm_kwargs,
}
with vllm_runner(**sd_args) as vllm_model:
sd_outputs = vllm_model.generate_w_logprobs(prompts, sampling_params)
num_returned_logprobs = [ num_returned_logprobs = [
len(logprob_dict) for seq_logprobs in spec_batch_logprobs len(seq_logprobs) for seq_logprobs in sd_outputs[-1]
for logprob_dict in seq_logprobs
] ]
# Assert one of the returned logprobs has > num_logprobs (indicating the # Assert one of the returned logprobs has > num_logprobs (indicating the
# sampled token is not in top-k). # sampled token is not in top-k).
assert any([ assert any(
num_returned > logprob_rank for num_returned in num_returned_logprobs [num_returned > logprobs for num_returned in num_returned_logprobs])
])
def run_greedy_logprobs_correctness_test(baseline_llm_generator,
test_llm_generator,
batch_size,
max_output_len,
force_output_len: bool,
logprob_rank: int = 1):
"""Helper method that compares the logprobs outputs of both the baseline LLM
and the test LLM. It asserts greedy equality of the logprobs when the
temperature is zero.
"""
temperature = 0.0
prompts = [
"Hello, my name is",
"The president of the United States is",
"The capital of France is",
"The future of AI is",
"San Francisco is know for its",
"Facebook was created in 2004 by",
"Curious George is a",
"Python 3.11 brings improvements to its",
]
prompts = [prompt for prompt, _ in zip(cycle(prompts), range(batch_size))]
# If the test requires that we generated max_output_len tokens, then set the
# sampling params to ignore eos token.
ignore_eos = force_output_len
sampling_params = SamplingParams(
max_tokens=max_output_len,
ignore_eos=ignore_eos,
temperature=temperature,
logprobs=logprob_rank,
)
spec_batch_logprobs = get_logprobs_from_llm_generator(
test_llm_generator, prompts, sampling_params)
baseline_batch_logprobs = get_logprobs_from_llm_generator(
baseline_llm_generator, prompts, sampling_params)
assert len(baseline_batch_logprobs) == len(prompts)
assert len(spec_batch_logprobs) == len(prompts)
# For each sequence in the batch.
for i, (baseline_logprobs, spec_logprobs) in enumerate(
zip(baseline_batch_logprobs, spec_batch_logprobs)):
assert len(spec_logprobs) == len(baseline_logprobs)
# For each generated position of the sequence.
for pos, (spec_pos_logprobs, baseline_pos_logprobs) in enumerate(
zip(spec_logprobs, baseline_logprobs)):
# Map rank to token/logprob in spec output.
spec_rank_to_token_id = {
value.rank: key
for key, value in spec_pos_logprobs.items()
}
spec_rank_to_logprob = {
value.rank: value.logprob
for key, value in spec_pos_logprobs.items()
}
# Map rank to token/logprob in baseline output.
baseline_rank_to_token_id = {
value.rank: key
for key, value in baseline_pos_logprobs.items()
}
baseline_rank_to_logprob = {
value.rank: value.logprob
for key, value in baseline_pos_logprobs.items()
}
# Assert set of ranks returned is equal.
assert set(spec_rank_to_token_id.keys()) == set(
baseline_rank_to_token_id.keys())
# Assert each logprob/token id is correct, keyed by rank.
for rank in sorted(set(spec_rank_to_token_id.keys())):
assert spec_rank_to_token_id[
rank] == baseline_rank_to_token_id[rank], f"{rank}"
assert math.isclose(
a=spec_rank_to_logprob[rank],
b=baseline_rank_to_logprob[rank],
abs_tol=1e-1,
)
@pytest.mark.parametrize( @pytest.mark.parametrize(
"common_llm_kwargs", "common_llm_kwargs",
[{ [{
"model": "JackFram/llama-160m", "model_name": "JackFram/llama-160m",
# Skip cuda graph recording for fast test. # Skip cuda graph recording for fast test.
"enforce_eager": True, "enforce_eager": True,
# Required for spec decode. # Required for spec decode.
"use_v2_block_manager": True, "use_v2_block_manager": True,
"max_logprobs": 6,
}]) }])
@pytest.mark.parametrize("per_test_common_llm_kwargs", [{}]) @pytest.mark.parametrize("per_test_common_llm_kwargs", [{}])
@pytest.mark.parametrize("baseline_llm_kwargs", [{}]) @pytest.mark.parametrize("baseline_llm_kwargs", [{}])
@ -364,57 +252,28 @@ def run_greedy_logprobs_correctness_test(baseline_llm_generator,
"disable_logprobs_during_spec_decoding": True, "disable_logprobs_during_spec_decoding": True,
}]) }])
@pytest.mark.parametrize("seed", [1]) @pytest.mark.parametrize("seed", [1])
def test_logprobs_disabled(baseline_llm_generator, test_llm_generator): @pytest.mark.parametrize("batch_size", [4])
@pytest.mark.parametrize(
"output_len",
[
# Use smaller output len for fast test.
32,
])
@pytest.mark.parametrize("logprobs", [0])
def test_logprobs_disabled(vllm_runner, common_llm_kwargs,
per_test_common_llm_kwargs, baseline_llm_kwargs,
test_llm_kwargs, batch_size: int, output_len: int,
seed: int, logprobs: int):
"""Check the behavior when logprobs are disabled. """Check the behavior when logprobs are disabled.
Token choices should match with the base model. Token choices should match with the base model.
""" """
prompts = [ run_logprob_correctness_test(vllm_runner,
"Hello, my name is", common_llm_kwargs,
"The president of the United States is", per_test_common_llm_kwargs,
"The capital of France is", baseline_llm_kwargs,
"The future of AI is", test_llm_kwargs,
"San Francisco is know for its", batch_size,
"Facebook was created in 2004 by", output_len,
"Curious George is a", seed,
"Python 3.11 brings improvements to its", temperature=0.0,
] logprobs=logprobs)
prompts = [prompt for prompt, _ in zip(cycle(prompts), range(4))]
sampling_params = SamplingParams(
# Use smaller output len for fast test
max_tokens=7,
ignore_eos=True,
temperature=0.0,
logprobs=2,
)
spec_batch_logprobs = get_logprobs_from_llm_generator(
test_llm_generator, prompts, sampling_params)
baseline_batch_logprobs = get_logprobs_from_llm_generator(
baseline_llm_generator, prompts, sampling_params)
assert len(baseline_batch_logprobs) == len(prompts)
assert len(spec_batch_logprobs) == len(prompts)
# For each sequence in the batch.
for _, (baseline_logprobs, spec_logprobs) in enumerate(
zip(baseline_batch_logprobs, spec_batch_logprobs)):
assert len(spec_logprobs) == len(baseline_logprobs)
# For each generated position of the sequence.
for _, (spec_pos_logprobs, baseline_pos_logprobs) in enumerate(
zip(spec_logprobs, baseline_logprobs)):
assert len(spec_pos_logprobs) == 1
spec_top_token_id = list(spec_pos_logprobs)[0]
spec_top_logprob = spec_pos_logprobs[spec_top_token_id]
assert spec_top_logprob.logprob == 0.0
assert spec_top_logprob.rank == -1
# check that the chosen token matches the base model
baseline_logprob = baseline_pos_logprobs[spec_top_token_id]
assert baseline_logprob.rank == 1
assert spec_top_logprob.decoded_token \
== baseline_logprob.decoded_token

118
tests/spec_decode/e2e/test_medusa_correctness.py
View File

@ -21,7 +21,7 @@ correctess for the target model outputs.
import pytest import pytest
from .conftest import run_greedy_equality_correctness_test from .conftest import run_equality_correctness_test
# main model # main model
# lmsys/vicuna-7b-v1.3 was to be used but it's causing # lmsys/vicuna-7b-v1.3 was to be used but it's causing
@ -55,7 +55,7 @@ PRECISION = "float32"
"dtype": PRECISION, "dtype": PRECISION,
# Main model # Main model
"model": MAIN_MODEL, "model_name": MAIN_MODEL,
}]) }])
@pytest.mark.parametrize("per_test_common_llm_kwargs", [{}]) @pytest.mark.parametrize("per_test_common_llm_kwargs", [{}])
@pytest.mark.parametrize("baseline_llm_kwargs", [{}]) @pytest.mark.parametrize("baseline_llm_kwargs", [{}])
@ -70,15 +70,21 @@ PRECISION = "float32"
]) ])
@pytest.mark.parametrize("batch_size", [1, 32]) @pytest.mark.parametrize("batch_size", [1, 32])
@pytest.mark.parametrize("seed", [1]) @pytest.mark.parametrize("seed", [1])
def test_medusa_e2e_greedy_correctness(baseline_llm_generator, def test_medusa_e2e_greedy_correctness(vllm_runner, common_llm_kwargs,
test_llm_generator, batch_size: int, per_test_common_llm_kwargs,
output_len: int): baseline_llm_kwargs, test_llm_kwargs,
batch_size: int, output_len: int,
seed: int):
"""Verify greedy equality with different batch size.""" """Verify greedy equality with different batch size."""
run_greedy_equality_correctness_test(baseline_llm_generator, run_equality_correctness_test(vllm_runner,
test_llm_generator, common_llm_kwargs,
batch_size, per_test_common_llm_kwargs,
max_output_len=output_len, baseline_llm_kwargs,
force_output_len=True) test_llm_kwargs,
batch_size,
max_output_len=output_len,
seed=seed,
temperature=0.0)
@pytest.mark.parametrize( @pytest.mark.parametrize(
@ -96,7 +102,7 @@ def test_medusa_e2e_greedy_correctness(baseline_llm_generator,
"dtype": PRECISION, "dtype": PRECISION,
# Main model # Main model
"model": MAIN_MODEL, "model_name": MAIN_MODEL,
}]) }])
@pytest.mark.parametrize("per_test_common_llm_kwargs", [{}]) @pytest.mark.parametrize("per_test_common_llm_kwargs", [{}])
@pytest.mark.parametrize("baseline_llm_kwargs", [{}]) @pytest.mark.parametrize("baseline_llm_kwargs", [{}])
@ -111,17 +117,21 @@ def test_medusa_e2e_greedy_correctness(baseline_llm_generator,
]) ])
@pytest.mark.parametrize("batch_size", [1, 32]) @pytest.mark.parametrize("batch_size", [1, 32])
@pytest.mark.parametrize("seed", [1]) @pytest.mark.parametrize("seed", [1])
def test_medusa_e2e_greedy_correctness_cuda_graph(baseline_llm_generator, def test_medusa_e2e_greedy_correctness_cuda_graph(
test_llm_generator, vllm_runner, common_llm_kwargs, per_test_common_llm_kwargs,
batch_size: int, baseline_llm_kwargs, test_llm_kwargs, batch_size: int, output_len: int,
output_len: int): seed: int):
"""Verify greedy equality with cuda graph enabled and different """Verify greedy equality with cuda graph enabled and different
batch sizes.""" batch sizes."""
run_greedy_equality_correctness_test(baseline_llm_generator, run_equality_correctness_test(vllm_runner,
test_llm_generator, common_llm_kwargs,
batch_size, per_test_common_llm_kwargs,
max_output_len=output_len, baseline_llm_kwargs,
force_output_len=True) test_llm_kwargs,
batch_size,
max_output_len=output_len,
seed=seed,
temperature=0.0)
@pytest.mark.parametrize( @pytest.mark.parametrize(
@ -142,7 +152,7 @@ def test_medusa_e2e_greedy_correctness_cuda_graph(baseline_llm_generator,
"dtype": PRECISION, "dtype": PRECISION,
# Main model # Main model
"model": MAIN_MODEL, "model_name": MAIN_MODEL,
}]) }])
@pytest.mark.parametrize("per_test_common_llm_kwargs", [{}]) @pytest.mark.parametrize("per_test_common_llm_kwargs", [{}])
@pytest.mark.parametrize("baseline_llm_kwargs", [{}]) @pytest.mark.parametrize("baseline_llm_kwargs", [{}])
@ -160,18 +170,22 @@ def test_medusa_e2e_greedy_correctness_cuda_graph(baseline_llm_generator,
]) ])
@pytest.mark.parametrize("batch_size", [4]) @pytest.mark.parametrize("batch_size", [4])
@pytest.mark.parametrize("seed", [1]) @pytest.mark.parametrize("seed", [1])
def test_medusa_e2e_greedy_correctness_with_preemption(baseline_llm_generator, def test_medusa_e2e_greedy_correctness_with_preemption(
test_llm_generator, vllm_runner, common_llm_kwargs, per_test_common_llm_kwargs,
batch_size: int, baseline_llm_kwargs, test_llm_kwargs, batch_size: int, output_len: int,
output_len: int): seed: int):
"""Verify greedy equality, even when some sequences are preempted mid- """Verify greedy equality, even when some sequences are preempted mid-
generation. generation.
""" """
run_greedy_equality_correctness_test(baseline_llm_generator, run_equality_correctness_test(vllm_runner,
test_llm_generator, common_llm_kwargs,
batch_size, per_test_common_llm_kwargs,
max_output_len=output_len, baseline_llm_kwargs,
force_output_len=True) test_llm_kwargs,
batch_size,
max_output_len=output_len,
seed=seed,
temperature=0.0)
@pytest.mark.parametrize( @pytest.mark.parametrize(
@ -187,7 +201,7 @@ def test_medusa_e2e_greedy_correctness_with_preemption(baseline_llm_generator,
"dtype": PRECISION, "dtype": PRECISION,
# Main model # Main model
"model": MAIN_MODEL, "model_name": MAIN_MODEL,
}]) }])
@pytest.mark.parametrize("per_test_common_llm_kwargs", [{}]) @pytest.mark.parametrize("per_test_common_llm_kwargs", [{}])
@pytest.mark.parametrize("baseline_llm_kwargs", [{}]) @pytest.mark.parametrize("baseline_llm_kwargs", [{}])
@ -209,16 +223,22 @@ def test_medusa_e2e_greedy_correctness_with_preemption(baseline_llm_generator,
32, 32,
]) ])
@pytest.mark.parametrize("seed", [1]) @pytest.mark.parametrize("seed", [1])
def test_medusa_different_k(baseline_llm_generator, test_llm_generator, def test_medusa_different_k(vllm_runner, common_llm_kwargs,
batch_size: int, output_len: int): per_test_common_llm_kwargs, baseline_llm_kwargs,
test_llm_kwargs, batch_size: int, output_len: int,
seed: int):
"""Verify that medusa speculative decoding produces exact equality """Verify that medusa speculative decoding produces exact equality
to without spec decode with different values of num_speculative_tokens. to without spec decode with different values of num_speculative_tokens.
""" """
run_greedy_equality_correctness_test(baseline_llm_generator, run_equality_correctness_test(vllm_runner,
test_llm_generator, common_llm_kwargs,
batch_size, per_test_common_llm_kwargs,
max_output_len=output_len, baseline_llm_kwargs,
force_output_len=True) test_llm_kwargs,
batch_size,
max_output_len=output_len,
seed=seed,
temperature=0.0)
@pytest.mark.parametrize( @pytest.mark.parametrize(
@ -234,7 +254,7 @@ def test_medusa_different_k(baseline_llm_generator, test_llm_generator,
"dtype": PRECISION, "dtype": PRECISION,
# Main model # Main model
"model": MAIN_MODEL, "model_name": MAIN_MODEL,
}]) }])
@pytest.mark.parametrize("per_test_common_llm_kwargs", [{}]) @pytest.mark.parametrize("per_test_common_llm_kwargs", [{}])
@pytest.mark.parametrize("baseline_llm_kwargs", [{}]) @pytest.mark.parametrize("baseline_llm_kwargs", [{}])
@ -252,17 +272,23 @@ def test_medusa_different_k(baseline_llm_generator, test_llm_generator,
32, 32,
]) ])
@pytest.mark.parametrize("seed", [1]) @pytest.mark.parametrize("seed", [1])
def test_medusa_disable_queue(baseline_llm_generator, test_llm_generator, def test_medusa_disable_queue(vllm_runner, common_llm_kwargs,
batch_size: int, output_len: int): per_test_common_llm_kwargs, baseline_llm_kwargs,
test_llm_kwargs, batch_size: int,
output_len: int, seed: int):
"""Verify that medusa speculative decoding produces exact equality """Verify that medusa speculative decoding produces exact equality
to without spec decode when speculation is disabled for large to without spec decode when speculation is disabled for large
batch sizes. batch sizes.
""" """
run_greedy_equality_correctness_test(baseline_llm_generator, run_equality_correctness_test(vllm_runner,
test_llm_generator, common_llm_kwargs,
batch_size, per_test_common_llm_kwargs,
max_output_len=output_len, baseline_llm_kwargs,
force_output_len=True) test_llm_kwargs,
batch_size,
max_output_len=output_len,
seed=seed,
temperature=0.0)
if __name__ == "__main__": if __name__ == "__main__":

167
tests/spec_decode/e2e/test_mlp_correctness.py
View File

@ -25,8 +25,7 @@ import pytest
from vllm.model_executor.layers.vocab_parallel_embedding import pad_vocab_size from vllm.model_executor.layers.vocab_parallel_embedding import pad_vocab_size
from .conftest import (run_equality_correctness_test, from .conftest import run_equality_correctness_test
run_greedy_equality_correctness_test)
# main model # main model
MAIN_MODEL = "JackFram/llama-160m" MAIN_MODEL = "JackFram/llama-160m"
@ -58,7 +57,7 @@ PRECISION = "float32"
"dtype": PRECISION, "dtype": PRECISION,
# Main model # Main model
"model": MAIN_MODEL, "model_name": MAIN_MODEL,
}]) }])
@pytest.mark.parametrize("per_test_common_llm_kwargs", [{}]) @pytest.mark.parametrize("per_test_common_llm_kwargs", [{}])
@pytest.mark.parametrize("baseline_llm_kwargs", [{}]) @pytest.mark.parametrize("baseline_llm_kwargs", [{}])
@ -72,14 +71,21 @@ PRECISION = "float32"
]) ])
@pytest.mark.parametrize("batch_size", [1, 32]) @pytest.mark.parametrize("batch_size", [1, 32])
@pytest.mark.parametrize("seed", [1]) @pytest.mark.parametrize("seed", [1])
def test_mlp_e2e_greedy_correctness(baseline_llm_generator, test_llm_generator, def test_mlp_e2e_greedy_correctness(vllm_runner, common_llm_kwargs,
batch_size: int, output_len: int): per_test_common_llm_kwargs,
baseline_llm_kwargs, test_llm_kwargs,
batch_size: int, output_len: int,
seed: int):
"""Verify greedy equality with different batch size.""" """Verify greedy equality with different batch size."""
run_greedy_equality_correctness_test(baseline_llm_generator, run_equality_correctness_test(vllm_runner,
test_llm_generator, common_llm_kwargs,
batch_size, per_test_common_llm_kwargs,
max_output_len=output_len, baseline_llm_kwargs,
force_output_len=True) test_llm_kwargs,
batch_size,
max_output_len=output_len,
seed=seed,
temperature=0.0)
@pytest.mark.parametrize( @pytest.mark.parametrize(
@ -98,7 +104,7 @@ def test_mlp_e2e_greedy_correctness(baseline_llm_generator, test_llm_generator,
"dtype": PRECISION, "dtype": PRECISION,
# Main model # Main model
"model": MAIN_MODEL, "model_name": MAIN_MODEL,
}]) }])
@pytest.mark.parametrize("per_test_common_llm_kwargs", [{}]) @pytest.mark.parametrize("per_test_common_llm_kwargs", [{}])
@pytest.mark.parametrize("baseline_llm_kwargs", [{}]) @pytest.mark.parametrize("baseline_llm_kwargs", [{}])
@ -110,17 +116,21 @@ def test_mlp_e2e_greedy_correctness(baseline_llm_generator, test_llm_generator,
@pytest.mark.parametrize("output_len", [2048]) @pytest.mark.parametrize("output_len", [2048])
@pytest.mark.parametrize("batch_size", [1, 32]) @pytest.mark.parametrize("batch_size", [1, 32])
@pytest.mark.parametrize("seed", [1]) @pytest.mark.parametrize("seed", [1])
def test_mlp_e2e_acceptance_rate(baseline_llm_generator, test_llm_generator, def test_mlp_e2e_acceptance_rate(vllm_runner, common_llm_kwargs,
batch_size: int, output_len: int): per_test_common_llm_kwargs,
baseline_llm_kwargs, test_llm_kwargs,
batch_size: int, output_len: int, seed: int):
"""Verify acceptance rate with different batch size and large output """Verify acceptance rate with different batch size and large output
length.""" length."""
run_equality_correctness_test(baseline_llm_generator, run_equality_correctness_test(vllm_runner,
test_llm_generator, common_llm_kwargs,
per_test_common_llm_kwargs,
baseline_llm_kwargs,
test_llm_kwargs,
batch_size, batch_size,
max_output_len=output_len, max_output_len=output_len,
temperature=0.0, temperature=0.0,
seeded=True, seed=seed,
force_output_len=True,
expected_acceptance_rate=0.48) expected_acceptance_rate=0.48)
@ -140,7 +150,7 @@ def test_mlp_e2e_acceptance_rate(baseline_llm_generator, test_llm_generator,
"dtype": PRECISION, "dtype": PRECISION,
# Main model # Main model
"model": MAIN_MODEL, "model_name": MAIN_MODEL,
# Speculative model # Speculative model
"speculative_model": SPEC_MODEL, "speculative_model": SPEC_MODEL,
@ -151,28 +161,35 @@ def test_mlp_e2e_acceptance_rate(baseline_llm_generator, test_llm_generator,
@pytest.mark.parametrize("output_len", [64]) @pytest.mark.parametrize("output_len", [64])
@pytest.mark.parametrize("batch_size", [1, 32]) @pytest.mark.parametrize("batch_size", [1, 32])
@pytest.mark.parametrize("temperature", [0.1, 1.0]) @pytest.mark.parametrize("temperature", [0.1, 1.0])
@pytest.mark.parametrize("seed", [None]) @pytest.mark.parametrize("seed", [1])
def test_mlp_e2e_seeded_correctness(baseline_llm_generator, test_llm_generator, def test_mlp_e2e_seeded_correctness(vllm_runner, common_llm_kwargs,
per_test_common_llm_kwargs,
baseline_llm_kwargs, test_llm_kwargs,
batch_size: int, output_len: int, batch_size: int, output_len: int,
temperature: float): temperature: float, seed: int):
"""Verify seeded runs produce the same output.""" """Verify seeded runs produce the same output."""
run_equality_correctness_test(baseline_llm_generator, run_equality_correctness_test(vllm_runner,
test_llm_generator, common_llm_kwargs,
per_test_common_llm_kwargs,
baseline_llm_kwargs,
test_llm_kwargs,
batch_size, batch_size,
max_output_len=output_len, max_output_len=output_len,
temperature=temperature, temperature=temperature,
seeded=True, seed=seed)
force_output_len=True)
# Ensure this same test does fail if we _don't_ include per-request seeds # Ensure this same test does fail if we _don't_ include per-request seeds
with pytest.raises(AssertionError): with pytest.raises(AssertionError):
run_equality_correctness_test(baseline_llm_generator, run_equality_correctness_test(vllm_runner,
test_llm_generator, common_llm_kwargs,
per_test_common_llm_kwargs,
baseline_llm_kwargs,
test_llm_kwargs,
batch_size, batch_size,
max_output_len=output_len, max_output_len=output_len,
temperature=temperature, temperature=temperature,
seeded=False, seed=seed,
force_output_len=True) disable_seed=True)
@pytest.mark.parametrize( @pytest.mark.parametrize(
@ -193,7 +210,7 @@ def test_mlp_e2e_seeded_correctness(baseline_llm_generator, test_llm_generator,
"dtype": PRECISION, "dtype": PRECISION,
# Main model # Main model
"model": MAIN_MODEL, "model_name": MAIN_MODEL,
}]) }])
@pytest.mark.parametrize("per_test_common_llm_kwargs", [{}]) @pytest.mark.parametrize("per_test_common_llm_kwargs", [{}])
@pytest.mark.parametrize("baseline_llm_kwargs", [{}]) @pytest.mark.parametrize("baseline_llm_kwargs", [{}])
@ -210,18 +227,22 @@ def test_mlp_e2e_seeded_correctness(baseline_llm_generator, test_llm_generator,
]) ])
@pytest.mark.parametrize("batch_size", [4]) @pytest.mark.parametrize("batch_size", [4])
@pytest.mark.parametrize("seed", [1]) @pytest.mark.parametrize("seed", [1])
def test_mlp_e2e_greedy_correctness_with_preemption(baseline_llm_generator, def test_mlp_e2e_greedy_correctness_with_preemption(
test_llm_generator, vllm_runner, common_llm_kwargs, per_test_common_llm_kwargs,
batch_size: int, baseline_llm_kwargs, test_llm_kwargs, batch_size: int, output_len: int,
output_len: int): seed: int):
"""Verify greedy equality, even when some sequences are preempted mid- """Verify greedy equality, even when some sequences are preempted mid-
generation. generation.
""" """
run_greedy_equality_correctness_test(baseline_llm_generator, run_equality_correctness_test(vllm_runner,
test_llm_generator, common_llm_kwargs,
batch_size, per_test_common_llm_kwargs,
max_output_len=output_len, baseline_llm_kwargs,
force_output_len=True) test_llm_kwargs,
batch_size,
max_output_len=output_len,
seed=seed,
temperature=0.0)
@pytest.mark.parametrize( @pytest.mark.parametrize(
@ -242,7 +263,7 @@ def test_mlp_e2e_greedy_correctness_with_preemption(baseline_llm_generator,
"dtype": PRECISION, "dtype": PRECISION,
# Main model # Main model
"model": MAIN_MODEL, "model_name": MAIN_MODEL,
}]) }])
@pytest.mark.parametrize("per_test_common_llm_kwargs", [{}]) @pytest.mark.parametrize("per_test_common_llm_kwargs", [{}])
@pytest.mark.parametrize("baseline_llm_kwargs", [{}]) @pytest.mark.parametrize("baseline_llm_kwargs", [{}])
@ -259,10 +280,10 @@ def test_mlp_e2e_greedy_correctness_with_preemption(baseline_llm_generator,
]) ])
@pytest.mark.parametrize("batch_size", [4]) @pytest.mark.parametrize("batch_size", [4])
@pytest.mark.parametrize("seed", [1]) @pytest.mark.parametrize("seed", [1])
def test_mlp_e2e_greedy_correctness_with_padding(baseline_llm_generator, def test_mlp_e2e_greedy_correctness_with_padding(
test_llm_generator, vllm_runner, common_llm_kwargs, per_test_common_llm_kwargs,
batch_size: int, baseline_llm_kwargs, test_llm_kwargs, batch_size: int, output_len: int,
output_len: int): seed: int):
"""Verify greedy equality when the vocab dimension is padded """Verify greedy equality when the vocab dimension is padded
""" """
@ -273,11 +294,15 @@ def test_mlp_e2e_greedy_correctness_with_padding(baseline_llm_generator,
with patch( with patch(
"vllm.model_executor.layers.vocab_parallel_embedding.pad_vocab_size", "vllm.model_executor.layers.vocab_parallel_embedding.pad_vocab_size",
patched_pad_vocab_size): patched_pad_vocab_size):
run_greedy_equality_correctness_test(baseline_llm_generator, run_equality_correctness_test(vllm_runner,
test_llm_generator, common_llm_kwargs,
batch_size, per_test_common_llm_kwargs,
max_output_len=output_len, baseline_llm_kwargs,
force_output_len=True) test_llm_kwargs,
batch_size,
max_output_len=output_len,
seed=seed,
temperature=0.0)
@pytest.mark.parametrize( @pytest.mark.parametrize(
@ -293,7 +318,7 @@ def test_mlp_e2e_greedy_correctness_with_padding(baseline_llm_generator,
"dtype": PRECISION, "dtype": PRECISION,
# Main model # Main model
"model": MAIN_MODEL, "model_name": MAIN_MODEL,
}]) }])
@pytest.mark.parametrize("per_test_common_llm_kwargs", [{}]) @pytest.mark.parametrize("per_test_common_llm_kwargs", [{}])
@pytest.mark.parametrize("baseline_llm_kwargs", [{}]) @pytest.mark.parametrize("baseline_llm_kwargs", [{}])
@ -315,16 +340,22 @@ def test_mlp_e2e_greedy_correctness_with_padding(baseline_llm_generator,
32, 32,
]) ])
@pytest.mark.parametrize("seed", [1]) @pytest.mark.parametrize("seed", [1])
def test_mlp_different_k(baseline_llm_generator, test_llm_generator, def test_mlp_different_k(vllm_runner, common_llm_kwargs,
batch_size: int, output_len: int): per_test_common_llm_kwargs, baseline_llm_kwargs,
test_llm_kwargs, batch_size: int, seed: int,
output_len: int):
"""Verify that mlp speculative decoding produces exact equality """Verify that mlp speculative decoding produces exact equality
to without spec decode with different values of num_speculative_tokens. to without spec decode with different values of num_speculative_tokens.
""" """
run_greedy_equality_correctness_test(baseline_llm_generator, run_equality_correctness_test(vllm_runner,
test_llm_generator, common_llm_kwargs,
batch_size, per_test_common_llm_kwargs,
max_output_len=output_len, baseline_llm_kwargs,
force_output_len=True) test_llm_kwargs,
batch_size,
max_output_len=output_len,
seed=seed,
temperature=0.0)
@pytest.mark.parametrize( @pytest.mark.parametrize(
@ -340,7 +371,7 @@ def test_mlp_different_k(baseline_llm_generator, test_llm_generator,
"dtype": PRECISION, "dtype": PRECISION,
# Main model # Main model
"model": MAIN_MODEL, "model_name": MAIN_MODEL,
}]) }])
@pytest.mark.parametrize("per_test_common_llm_kwargs", [{}]) @pytest.mark.parametrize("per_test_common_llm_kwargs", [{}])
@pytest.mark.parametrize("baseline_llm_kwargs", [{}]) @pytest.mark.parametrize("baseline_llm_kwargs", [{}])
@ -357,14 +388,20 @@ def test_mlp_different_k(baseline_llm_generator, test_llm_generator,
32, 32,
]) ])
@pytest.mark.parametrize("seed", [1]) @pytest.mark.parametrize("seed", [1])
def test_mlp_disable_queue(baseline_llm_generator, test_llm_generator, def test_mlp_disable_queue(vllm_runner, common_llm_kwargs,
batch_size: int, output_len: int): per_test_common_llm_kwargs, baseline_llm_kwargs,
test_llm_kwargs, batch_size: int, seed: int,
output_len: int):
"""Verify that mlp speculative decoding produces exact equality """Verify that mlp speculative decoding produces exact equality
to without spec decode when speculation is disabled for large to without spec decode when speculation is disabled for large
batch sizes. batch sizes.
""" """
run_greedy_equality_correctness_test(baseline_llm_generator, run_equality_correctness_test(vllm_runner,
test_llm_generator, common_llm_kwargs,
batch_size, per_test_common_llm_kwargs,
max_output_len=output_len, baseline_llm_kwargs,
force_output_len=True) test_llm_kwargs,
batch_size,
max_output_len=output_len,
seed=seed,
temperature=0.0)

307
tests/spec_decode/e2e/test_multistep_correctness.py
View File

@ -41,8 +41,9 @@ from transformers import AutoTokenizer
from vllm import SamplingParams from vllm import SamplingParams
from ...utils import fork_new_process_for_each_test
from .conftest import (get_output_from_llm_generator, from .conftest import (get_output_from_llm_generator,
run_greedy_equality_correctness_test) run_equality_correctness_test)
@pytest.mark.parametrize( @pytest.mark.parametrize(
@ -73,6 +74,7 @@ from .conftest import (get_output_from_llm_generator,
@pytest.mark.parametrize("test_llm_kwargs", [{}]) @pytest.mark.parametrize("test_llm_kwargs", [{}])
@pytest.mark.parametrize("batch_size", [1, 32]) @pytest.mark.parametrize("batch_size", [1, 32])
@pytest.mark.parametrize("seed", [1]) @pytest.mark.parametrize("seed", [1])
@fork_new_process_for_each_test
def test_spec_decode_e2e_with_detokenization(test_llm_generator, def test_spec_decode_e2e_with_detokenization(test_llm_generator,
batch_size: int): batch_size: int):
"""Run generation with speculative decoding on a batch. Verify the engine """Run generation with speculative decoding on a batch. Verify the engine
@ -116,44 +118,6 @@ def test_spec_decode_e2e_with_detokenization(test_llm_generator,
assert actual_tokens.strip() == expected_tokens.strip() assert actual_tokens.strip() == expected_tokens.strip()
@pytest.mark.parametrize(
"common_llm_kwargs",
[{
# Use a small model for a fast test.
# Note this is repeated in the test body; to initialize a tokenizer.
"model": "JackFram/llama-68m",
# Skip cuda graph recording for fast test.
"enforce_eager": True,
# Required for spec decode.
"use_v2_block_manager": True,
# Use AsyncLLM engine
"use_async": True,
}])
@pytest.mark.parametrize("baseline_llm_kwargs", [{}])
@pytest.mark.parametrize("per_test_common_llm_kwargs", [
{
"speculative_model": "JackFram/llama-68m",
"num_speculative_tokens": 5,
},
])
@pytest.mark.parametrize("test_llm_kwargs", [{}])
@pytest.mark.parametrize("batch_size", [2])
@pytest.mark.parametrize("seed", [1])
def test_spec_decode_e2e_with_async_engine(test_llm_generator,
baseline_llm_generator,
batch_size: int):
"""Verify spec decode works well with async LLM engine.
"""
run_greedy_equality_correctness_test(baseline_llm_generator,
test_llm_generator,
batch_size,
max_output_len=32,
force_output_len=True)
@pytest.mark.parametrize( @pytest.mark.parametrize(
"common_llm_kwargs", "common_llm_kwargs",
[{ [{
@ -172,10 +136,10 @@ def test_spec_decode_e2e_with_async_engine(test_llm_generator,
# Try two different tiny base models. # Try two different tiny base models.
# Note that one is equal to the draft model, another isn't. # Note that one is equal to the draft model, another isn't.
{ {
"model": "JackFram/llama-68m", "model_name": "JackFram/llama-68m",
}, },
{ {
"model": "JackFram/llama-160m", "model_name": "JackFram/llama-160m",
}, },
]) ])
@pytest.mark.parametrize("baseline_llm_kwargs", [{}]) @pytest.mark.parametrize("baseline_llm_kwargs", [{}])
@ -189,13 +153,15 @@ def test_spec_decode_e2e_with_async_engine(test_llm_generator,
"output_len", "output_len",
[ [
# Use long output len for the small model test. # Use long output len for the small model test.
1536, 10,
]) ])
@pytest.mark.parametrize("batch_size", [1]) @pytest.mark.parametrize("batch_size", [1])
@pytest.mark.parametrize("seed", [1]) @pytest.mark.parametrize("seed", [1])
@fork_new_process_for_each_test
def test_spec_decode_e2e_greedy_correctness_tiny_model_bs1( def test_spec_decode_e2e_greedy_correctness_tiny_model_bs1(
baseline_llm_generator, test_llm_generator, batch_size: int, vllm_runner, common_llm_kwargs, per_test_common_llm_kwargs,
output_len: int): baseline_llm_kwargs, test_llm_kwargs, batch_size: int, output_len: int,
seed: int):
"""Verify greedy equality on a tiny model with batch size of one. """Verify greedy equality on a tiny model with batch size of one.
Since this test is cheaper than other e2e correctness tests, we generate Since this test is cheaper than other e2e correctness tests, we generate
@ -204,14 +170,18 @@ def test_spec_decode_e2e_greedy_correctness_tiny_model_bs1(
When the draft model is the same as the target model, we further check When the draft model is the same as the target model, we further check
whether all speculative tokens are accepted. whether all speculative tokens are accepted.
""" """
ensure_all_accepted = test_llm_generator.same_draft_target_model ensure_all_accepted = per_test_common_llm_kwargs.get(
run_greedy_equality_correctness_test( "model_name") == test_llm_kwargs.get("speculative_model")
baseline_llm_generator, run_equality_correctness_test(vllm_runner,
test_llm_generator, common_llm_kwargs,
batch_size, per_test_common_llm_kwargs,
max_output_len=output_len, baseline_llm_kwargs,
force_output_len=True, test_llm_kwargs,
ensure_all_accepted=ensure_all_accepted) batch_size,
max_output_len=output_len,
seed=seed,
temperature=0.0,
ensure_all_accepted=ensure_all_accepted)
@pytest.mark.parametrize( @pytest.mark.parametrize(
@ -232,10 +202,10 @@ def test_spec_decode_e2e_greedy_correctness_tiny_model_bs1(
# Try two different tiny base models. # Try two different tiny base models.
# Note that one is equal to the draft model, another isn't. # Note that one is equal to the draft model, another isn't.
{ {
"model": "JackFram/llama-68m", "model_name": "JackFram/llama-68m",
}, },
{ {
"model": "JackFram/llama-160m", "model_name": "JackFram/llama-160m",
}, },
]) ])
@pytest.mark.parametrize("baseline_llm_kwargs", [{}]) @pytest.mark.parametrize("baseline_llm_kwargs", [{}])
@ -253,16 +223,22 @@ def test_spec_decode_e2e_greedy_correctness_tiny_model_bs1(
]) ])
@pytest.mark.parametrize("batch_size", [64]) @pytest.mark.parametrize("batch_size", [64])
@pytest.mark.parametrize("seed", [1]) @pytest.mark.parametrize("seed", [1])
@fork_new_process_for_each_test
def test_spec_decode_e2e_greedy_correctness_tiny_model_large_bs( def test_spec_decode_e2e_greedy_correctness_tiny_model_large_bs(
baseline_llm_generator, test_llm_generator, batch_size: int, vllm_runner, common_llm_kwargs, per_test_common_llm_kwargs,
output_len: int): baseline_llm_kwargs, test_llm_kwargs, batch_size: int, output_len: int,
seed: int):
"""Verify greedy equality on a tiny model and large batch size. """Verify greedy equality on a tiny model and large batch size.
""" """
run_greedy_equality_correctness_test(baseline_llm_generator, run_equality_correctness_test(vllm_runner,
test_llm_generator, common_llm_kwargs,
batch_size, per_test_common_llm_kwargs,
max_output_len=output_len, baseline_llm_kwargs,
force_output_len=True) test_llm_kwargs,
batch_size,
max_output_len=output_len,
seed=seed,
temperature=0.0)
@pytest.mark.parametrize( @pytest.mark.parametrize(
@ -280,10 +256,10 @@ def test_spec_decode_e2e_greedy_correctness_tiny_model_large_bs(
# Try two different tiny base models. # Try two different tiny base models.
# Note that one is equal to the draft model, another isn't. # Note that one is equal to the draft model, another isn't.
{ {
"model": "JackFram/llama-68m", "model_name": "JackFram/llama-68m",
}, },
{ {
"model": "JackFram/llama-160m", "model_name": "JackFram/llama-160m",
}, },
]) ])
@pytest.mark.parametrize("baseline_llm_kwargs", [{}]) @pytest.mark.parametrize("baseline_llm_kwargs", [{}])
@ -298,24 +274,31 @@ def test_spec_decode_e2e_greedy_correctness_tiny_model_large_bs(
]) ])
@pytest.mark.parametrize("batch_size", [32]) @pytest.mark.parametrize("batch_size", [32])
@pytest.mark.parametrize("seed", [1]) @pytest.mark.parametrize("seed", [1])
@fork_new_process_for_each_test
def test_spec_decode_e2e_greedy_correctness_tiny_model_large_bs_diff_output_len( def test_spec_decode_e2e_greedy_correctness_tiny_model_large_bs_diff_output_len(
baseline_llm_generator, test_llm_generator, batch_size: int, vllm_runner, common_llm_kwargs, per_test_common_llm_kwargs,
max_output_len: int): baseline_llm_kwargs, test_llm_kwargs, batch_size: int,
max_output_len: int, seed: int):
"""Verify greedy equality on a tiny model, with a large batch size, and when """Verify greedy equality on a tiny model, with a large batch size, and when
sampling respects the EOS token. sampling respects the EOS token.
""" """
run_greedy_equality_correctness_test(baseline_llm_generator, run_equality_correctness_test(vllm_runner,
test_llm_generator, common_llm_kwargs,
batch_size, per_test_common_llm_kwargs,
max_output_len, baseline_llm_kwargs,
force_output_len=False) test_llm_kwargs,
batch_size,
max_output_len,
seed=seed,
temperature=0.0,
ignore_eos=False)
@pytest.mark.parametrize( @pytest.mark.parametrize(
"common_llm_kwargs", "common_llm_kwargs",
[{ [{
# A "real" model (not tiny). # A "real" model (not tiny).
"model": "meta-llama/Llama-2-7b-chat-hf", "model_name": "meta-llama/Llama-2-7b-chat-hf",
# Skip cuda graph recording for fast test. # Skip cuda graph recording for fast test.
"enforce_eager": True, "enforce_eager": True,
@ -342,24 +325,30 @@ def test_spec_decode_e2e_greedy_correctness_tiny_model_large_bs_diff_output_len(
256, 256,
]) ])
@pytest.mark.parametrize("seed", [1]) @pytest.mark.parametrize("seed", [1])
@fork_new_process_for_each_test
def test_spec_decode_e2e_greedy_correctness_real_model_bs1( def test_spec_decode_e2e_greedy_correctness_real_model_bs1(
baseline_llm_generator, test_llm_generator, batch_size: int, vllm_runner, common_llm_kwargs, per_test_common_llm_kwargs,
output_len: int): baseline_llm_kwargs, test_llm_kwargs, batch_size: int, output_len: int,
seed: int):
"""Verify greedy equality on a "real" model and batch size of 1. This is """Verify greedy equality on a "real" model and batch size of 1. This is
separate from large BS tests to make identifying the source of bugs easier. separate from large BS tests to make identifying the source of bugs easier.
""" """
run_greedy_equality_correctness_test(baseline_llm_generator, run_equality_correctness_test(vllm_runner,
test_llm_generator, common_llm_kwargs,
batch_size, per_test_common_llm_kwargs,
max_output_len=output_len, baseline_llm_kwargs,
force_output_len=True) test_llm_kwargs,
batch_size,
max_output_len=output_len,
seed=seed,
temperature=0.0)
@pytest.mark.parametrize( @pytest.mark.parametrize(
"common_llm_kwargs", "common_llm_kwargs",
[{ [{
# A "real" model (not tiny). # A "real" model (not tiny).
"model": "meta-llama/Llama-2-7b-chat-hf", "model_name": "meta-llama/Llama-2-7b-chat-hf",
# Skip cuda graph recording for fast test. # Skip cuda graph recording for fast test.
"enforce_eager": True, "enforce_eager": True,
@ -386,17 +375,23 @@ def test_spec_decode_e2e_greedy_correctness_real_model_bs1(
64, 64,
]) ])
@pytest.mark.parametrize("seed", [1]) @pytest.mark.parametrize("seed", [1])
@fork_new_process_for_each_test
def test_spec_decode_e2e_greedy_correctness_real_model_large_bs( def test_spec_decode_e2e_greedy_correctness_real_model_large_bs(
baseline_llm_generator, test_llm_generator, batch_size: int, vllm_runner, common_llm_kwargs, per_test_common_llm_kwargs,
output_len: int): baseline_llm_kwargs, test_llm_kwargs, batch_size: int, output_len: int,
seed: int):
"""Verify greedy equality with a "real" model on a nontrivial batch size. """Verify greedy equality with a "real" model on a nontrivial batch size.
This is the closest test to a real production workload. This is the closest test to a real production workload.
""" """
run_greedy_equality_correctness_test(baseline_llm_generator, run_equality_correctness_test(vllm_runner,
test_llm_generator, common_llm_kwargs,
batch_size, per_test_common_llm_kwargs,
max_output_len=output_len, baseline_llm_kwargs,
force_output_len=True) test_llm_kwargs,
batch_size,
max_output_len=output_len,
seed=seed,
temperature=0.0)
@pytest.mark.parametrize( @pytest.mark.parametrize(
@ -415,7 +410,7 @@ def test_spec_decode_e2e_greedy_correctness_real_model_large_bs(
}]) }])
@pytest.mark.parametrize("per_test_common_llm_kwargs", [ @pytest.mark.parametrize("per_test_common_llm_kwargs", [
{ {
"model": "JackFram/llama-160m", "model_name": "JackFram/llama-160m",
}, },
]) ])
@pytest.mark.parametrize("baseline_llm_kwargs", [{}]) @pytest.mark.parametrize("baseline_llm_kwargs", [{}])
@ -433,23 +428,29 @@ def test_spec_decode_e2e_greedy_correctness_real_model_large_bs(
]) ])
@pytest.mark.parametrize("batch_size", [4]) @pytest.mark.parametrize("batch_size", [4])
@pytest.mark.parametrize("seed", [1]) @pytest.mark.parametrize("seed", [1])
@fork_new_process_for_each_test
def test_spec_decode_e2e_greedy_correctness_with_preemption( def test_spec_decode_e2e_greedy_correctness_with_preemption(
baseline_llm_generator, test_llm_generator, batch_size: int, vllm_runner, common_llm_kwargs, per_test_common_llm_kwargs,
output_len: int): baseline_llm_kwargs, test_llm_kwargs, batch_size: int, output_len: int,
seed: int):
"""Verify greedy equality, even when some sequences are preempted mid- """Verify greedy equality, even when some sequences are preempted mid-
generation. generation.
""" """
run_greedy_equality_correctness_test(baseline_llm_generator, run_equality_correctness_test(vllm_runner,
test_llm_generator, common_llm_kwargs,
batch_size, per_test_common_llm_kwargs,
max_output_len=output_len, baseline_llm_kwargs,
force_output_len=True) test_llm_kwargs,
batch_size,
max_output_len=output_len,
seed=seed,
temperature=0.0)
@pytest.mark.parametrize( @pytest.mark.parametrize(
"common_llm_kwargs", "common_llm_kwargs",
[{ [{
"model": "JackFram/llama-160m", "model_name": "JackFram/llama-160m",
# Skip cuda graph recording for fast test. # Skip cuda graph recording for fast test.
"enforce_eager": True, "enforce_eager": True,
@ -487,22 +488,29 @@ def test_spec_decode_e2e_greedy_correctness_with_preemption(
32, 32,
]) ])
@pytest.mark.parametrize("seed", [1]) @pytest.mark.parametrize("seed", [1])
def test_spec_decode_different_block_size(baseline_llm_generator, @fork_new_process_for_each_test
test_llm_generator, batch_size: int, def test_spec_decode_different_block_size(vllm_runner, common_llm_kwargs,
output_len: int): per_test_common_llm_kwargs,
baseline_llm_kwargs, test_llm_kwargs,
batch_size: int, output_len: int,
seed: int):
"""Verify greedy equality over different block sizes. """Verify greedy equality over different block sizes.
""" """
run_greedy_equality_correctness_test(baseline_llm_generator, run_equality_correctness_test(vllm_runner,
test_llm_generator, common_llm_kwargs,
batch_size, per_test_common_llm_kwargs,
max_output_len=output_len, baseline_llm_kwargs,
force_output_len=True) test_llm_kwargs,
batch_size,
max_output_len=output_len,
seed=seed,
temperature=0.0)
@pytest.mark.parametrize( @pytest.mark.parametrize(
"common_llm_kwargs", "common_llm_kwargs",
[{ [{
"model": "JackFram/llama-160m", "model_name": "JackFram/llama-160m",
# Skip cuda graph recording for fast test. # Skip cuda graph recording for fast test.
"enforce_eager": True, "enforce_eager": True,
@ -534,24 +542,31 @@ def test_spec_decode_different_block_size(baseline_llm_generator,
64, 64,
]) ])
@pytest.mark.parametrize("seed", [1]) @pytest.mark.parametrize("seed", [1])
def test_skip_speculation(baseline_llm_generator, test_llm_generator, @fork_new_process_for_each_test
batch_size: int, output_len: int): def test_skip_speculation(vllm_runner, common_llm_kwargs,
per_test_common_llm_kwargs, baseline_llm_kwargs,
test_llm_kwargs, batch_size: int, output_len: int,
seed: int):
"""Verify greedy equality when some (or all) sequences skip speculation. """Verify greedy equality when some (or all) sequences skip speculation.
We do this by setting the max model len of the draft model to an We do this by setting the max model len of the draft model to an
artificially low value, such that when the sequences grow beyond it, they artificially low value, such that when the sequences grow beyond it, they
are skipped in speculative decoding. are skipped in speculative decoding.
""" """
run_greedy_equality_correctness_test(baseline_llm_generator, run_equality_correctness_test(vllm_runner,
test_llm_generator, common_llm_kwargs,
batch_size, per_test_common_llm_kwargs,
max_output_len=output_len, baseline_llm_kwargs,
force_output_len=True) test_llm_kwargs,
batch_size,
max_output_len=output_len,
seed=seed,
temperature=0.0)
@pytest.mark.parametrize( @pytest.mark.parametrize(
"common_llm_kwargs", "common_llm_kwargs",
[{ [{
"model": "JackFram/llama-160m", "model_name": "JackFram/llama-160m",
# Skip cuda graph recording for fast test. # Skip cuda graph recording for fast test.
"enforce_eager": True, "enforce_eager": True,
@ -571,21 +586,28 @@ def test_skip_speculation(baseline_llm_generator, test_llm_generator,
@pytest.mark.parametrize("batch_size", [8]) @pytest.mark.parametrize("batch_size", [8])
@pytest.mark.parametrize("output_len", [10]) @pytest.mark.parametrize("output_len", [10])
@pytest.mark.parametrize("seed", [1]) @pytest.mark.parametrize("seed", [1])
def test_disable_speculation(baseline_llm_generator, test_llm_generator, @fork_new_process_for_each_test
batch_size: int, output_len: int): def test_disable_speculation(vllm_runner, common_llm_kwargs,
per_test_common_llm_kwargs, baseline_llm_kwargs,
test_llm_kwargs, batch_size: int, output_len: int,
seed: int):
"""Verify greedy equality when all sequences disable speculation. """Verify greedy equality when all sequences disable speculation.
""" """
run_greedy_equality_correctness_test(baseline_llm_generator, run_equality_correctness_test(vllm_runner,
test_llm_generator, common_llm_kwargs,
batch_size, per_test_common_llm_kwargs,
max_output_len=output_len, baseline_llm_kwargs,
force_output_len=True) test_llm_kwargs,
batch_size,
max_output_len=output_len,
seed=seed,
temperature=0.0)
@pytest.mark.parametrize( @pytest.mark.parametrize(
"common_llm_kwargs", "common_llm_kwargs",
[{ [{
"model": "JackFram/llama-68m", "model_name": "JackFram/llama-68m",
# Skip cuda graph recording for fast test. # Skip cuda graph recording for fast test.
"enforce_eager": True, "enforce_eager": True,
@ -613,22 +635,28 @@ def test_disable_speculation(baseline_llm_generator, test_llm_generator,
32, 32,
]) ])
@pytest.mark.parametrize("seed", [1]) @pytest.mark.parametrize("seed", [1])
def test_many_k(baseline_llm_generator, test_llm_generator, batch_size: int, @fork_new_process_for_each_test
output_len: int): def test_many_k(vllm_runner, common_llm_kwargs, per_test_common_llm_kwargs,
baseline_llm_kwargs, test_llm_kwargs, batch_size: int,
output_len: int, seed: int):
"""Verify that speculative decoding produces exact equality to without spec """Verify that speculative decoding produces exact equality to without spec
decode with many different values of k. decode with many different values of k.
""" """
run_greedy_equality_correctness_test(baseline_llm_generator, run_equality_correctness_test(vllm_runner,
test_llm_generator, common_llm_kwargs,
batch_size, per_test_common_llm_kwargs,
max_output_len=output_len, baseline_llm_kwargs,
force_output_len=True) test_llm_kwargs,
batch_size,
max_output_len=output_len,
seed=seed,
temperature=0.0)
@pytest.mark.parametrize( @pytest.mark.parametrize(
"common_llm_kwargs", "common_llm_kwargs",
[{ [{
"model": "JackFram/llama-160m", "model_name": "JackFram/llama-160m",
# Skip cuda graph recording for fast test. # Skip cuda graph recording for fast test.
"enforce_eager": True, "enforce_eager": True,
@ -657,15 +685,22 @@ def test_many_k(baseline_llm_generator, test_llm_generator, batch_size: int,
32, 32,
]) ])
@pytest.mark.parametrize("seed", [1]) @pytest.mark.parametrize("seed", [1])
def test_typical_acceptance_sampling(baseline_llm_generator, @fork_new_process_for_each_test
test_llm_generator, batch_size: int, def test_typical_acceptance_sampling(vllm_runner, common_llm_kwargs,
output_len: int): per_test_common_llm_kwargs,
baseline_llm_kwargs, test_llm_kwargs,
batch_size: int, output_len: int,
seed: int):
"""Verify that speculative decoding produces exact equality to without spec """Verify that speculative decoding produces exact equality to without spec
decode with TypicalAcceptanceSampler as the draft token acceptance decode with TypicalAcceptanceSampler as the draft token acceptance
sampling method. sampling method.
""" """
run_greedy_equality_correctness_test(baseline_llm_generator, run_equality_correctness_test(vllm_runner,
test_llm_generator, common_llm_kwargs,
batch_size, per_test_common_llm_kwargs,
max_output_len=output_len, baseline_llm_kwargs,
force_output_len=True) test_llm_kwargs,
batch_size,
max_output_len=output_len,
seed=seed,
temperature=0.0)

94
tests/spec_decode/e2e/test_ngram_correctness.py
View File

@ -26,7 +26,7 @@ for the target model outputs.
import pytest import pytest
from .conftest import run_greedy_equality_correctness_test from .conftest import run_equality_correctness_test
@pytest.mark.parametrize( @pytest.mark.parametrize(
@ -43,7 +43,7 @@ from .conftest import run_greedy_equality_correctness_test
}]) }])
@pytest.mark.parametrize("per_test_common_llm_kwargs", [ @pytest.mark.parametrize("per_test_common_llm_kwargs", [
{ {
"model": "JackFram/llama-68m", "model_name": "JackFram/llama-68m",
}, },
]) ])
@pytest.mark.parametrize("baseline_llm_kwargs", [{}]) @pytest.mark.parametrize("baseline_llm_kwargs", [{}])
@ -59,15 +59,21 @@ from .conftest import run_greedy_equality_correctness_test
]) ])
@pytest.mark.parametrize("batch_size", [1, 32]) @pytest.mark.parametrize("batch_size", [1, 32])
@pytest.mark.parametrize("seed", [1]) @pytest.mark.parametrize("seed", [1])
def test_ngram_e2e_greedy_correctness(baseline_llm_generator, def test_ngram_e2e_greedy_correctness(vllm_runner, common_llm_kwargs,
test_llm_generator, batch_size: int, per_test_common_llm_kwargs,
output_len: int): baseline_llm_kwargs, test_llm_kwargs,
batch_size: int, output_len: int,
seed: int):
"""Verify greedy equality on a tiny model with different batch size.""" """Verify greedy equality on a tiny model with different batch size."""
run_greedy_equality_correctness_test(baseline_llm_generator, run_equality_correctness_test(vllm_runner,
test_llm_generator, common_llm_kwargs,
batch_size, per_test_common_llm_kwargs,
max_output_len=output_len, baseline_llm_kwargs,
force_output_len=True) test_llm_kwargs,
batch_size,
max_output_len=output_len,
seed=seed,
temperature=0.0)
@pytest.mark.parametrize( @pytest.mark.parametrize(
@ -86,7 +92,7 @@ def test_ngram_e2e_greedy_correctness(baseline_llm_generator,
}]) }])
@pytest.mark.parametrize("per_test_common_llm_kwargs", [ @pytest.mark.parametrize("per_test_common_llm_kwargs", [
{ {
"model": "JackFram/llama-160m", "model_name": "JackFram/llama-160m",
}, },
]) ])
@pytest.mark.parametrize("baseline_llm_kwargs", [{}]) @pytest.mark.parametrize("baseline_llm_kwargs", [{}])
@ -105,24 +111,28 @@ def test_ngram_e2e_greedy_correctness(baseline_llm_generator,
]) ])
@pytest.mark.parametrize("batch_size", [4]) @pytest.mark.parametrize("batch_size", [4])
@pytest.mark.parametrize("seed", [1]) @pytest.mark.parametrize("seed", [1])
def test_ngram_e2e_greedy_correctness_with_preemption(baseline_llm_generator, def test_ngram_e2e_greedy_correctness_with_preemption(
test_llm_generator, vllm_runner, common_llm_kwargs, per_test_common_llm_kwargs,
batch_size: int, baseline_llm_kwargs, test_llm_kwargs, batch_size: int, output_len: int,
output_len: int): seed: int):
"""Verify greedy equality, even when some sequences are preempted mid- """Verify greedy equality, even when some sequences are preempted mid-
generation. generation.
""" """
run_greedy_equality_correctness_test(baseline_llm_generator, run_equality_correctness_test(vllm_runner,
test_llm_generator, common_llm_kwargs,
batch_size, per_test_common_llm_kwargs,
max_output_len=output_len, baseline_llm_kwargs,
force_output_len=True) test_llm_kwargs,
batch_size,
max_output_len=output_len,
temperature=0,
seed=seed)
@pytest.mark.parametrize( @pytest.mark.parametrize(
"common_llm_kwargs", "common_llm_kwargs",
[{ [{
"model": "JackFram/llama-68m", "model_name": "JackFram/llama-68m",
# Skip cuda graph recording for fast test. # Skip cuda graph recording for fast test.
"enforce_eager": True, "enforce_eager": True,
@ -159,23 +169,29 @@ def test_ngram_e2e_greedy_correctness_with_preemption(baseline_llm_generator,
32, 32,
]) ])
@pytest.mark.parametrize("seed", [1]) @pytest.mark.parametrize("seed", [1])
def test_ngram_different_k(baseline_llm_generator, test_llm_generator, def test_ngram_different_k(vllm_runner, common_llm_kwargs,
batch_size: int, output_len: int): per_test_common_llm_kwargs, baseline_llm_kwargs,
test_llm_kwargs, batch_size: int, output_len: int,
seed: int):
"""Verify that ngram speculative decoding produces exact equality """Verify that ngram speculative decoding produces exact equality
to without spec decode with many different values of k and to without spec decode with many different values of k and
different ngram_prompt_lookup_max. different ngram_prompt_lookup_max.
""" """
run_greedy_equality_correctness_test(baseline_llm_generator, run_equality_correctness_test(vllm_runner,
test_llm_generator, common_llm_kwargs,
batch_size, per_test_common_llm_kwargs,
max_output_len=output_len, baseline_llm_kwargs,
force_output_len=True) test_llm_kwargs,
batch_size,
max_output_len=output_len,
seed=seed,
temperature=0.0)
@pytest.mark.parametrize( @pytest.mark.parametrize(
"common_llm_kwargs", "common_llm_kwargs",
[{ [{
"model": "JackFram/llama-68m", "model_name": "JackFram/llama-68m",
# Skip cuda graph recording for fast test. # Skip cuda graph recording for fast test.
"enforce_eager": True, "enforce_eager": True,
@ -200,14 +216,20 @@ def test_ngram_different_k(baseline_llm_generator, test_llm_generator,
32, 32,
]) ])
@pytest.mark.parametrize("seed", [1]) @pytest.mark.parametrize("seed", [1])
def test_ngram_disable_queue(baseline_llm_generator, test_llm_generator, def test_ngram_disable_queue(vllm_runner, common_llm_kwargs,
batch_size: int, output_len: int): per_test_common_llm_kwargs, baseline_llm_kwargs,
test_llm_kwargs, batch_size: int, output_len: int,
seed: int):
"""Verify that ngram speculative decoding produces exact equality """Verify that ngram speculative decoding produces exact equality
to without spec decode with many different values of k and to without spec decode with many different values of k and
different ngram_prompt_lookup_max. different ngram_prompt_lookup_max.
""" """
run_greedy_equality_correctness_test(baseline_llm_generator, run_equality_correctness_test(vllm_runner,
test_llm_generator, common_llm_kwargs,
batch_size, per_test_common_llm_kwargs,
max_output_len=output_len, baseline_llm_kwargs,
force_output_len=True) test_llm_kwargs,
batch_size,
max_output_len=output_len,
seed=seed,
temperature=0.0)

52
tests/spec_decode/e2e/test_seed.py
View File

@ -2,11 +2,17 @@ import pytest
from .conftest import run_equality_correctness_test from .conftest import run_equality_correctness_test
# main model
MAIN_MODEL = "JackFram/llama-68m"
# speculative model
SPEC_MODEL = "JackFram/llama-160m"
@pytest.mark.parametrize( @pytest.mark.parametrize(
"common_llm_kwargs", "common_llm_kwargs",
[{ [{
"model": "JackFram/llama-68m", "model_name": "JackFram/llama-68m",
# Skip cuda graph recording for fast test. # Skip cuda graph recording for fast test.
"enforce_eager": True, "enforce_eager": True,
@ -31,26 +37,34 @@ from .conftest import run_equality_correctness_test
# Use smaller output len for fast test. # Use smaller output len for fast test.
20, 20,
]) ])
@pytest.mark.parametrize("seed", [None]) def test_seeded_consistency(vllm_runner, common_llm_kwargs,
def test_seeded_consistency(baseline_llm_generator, test_llm_generator, per_test_common_llm_kwargs, baseline_llm_kwargs,
batch_size: int, temperature: float, test_llm_kwargs, batch_size: int,
output_len: int): temperature: float, output_len: int):
"""Verify outputs are consistent across multiple runs with same seed """Verify outputs are consistent across multiple runs with same seed
""" """
run_equality_correctness_test(baseline_llm_generator, run_equality_correctness_test(
test_llm_generator, vllm_runner,
batch_size, common_llm_kwargs,
max_output_len=output_len, per_test_common_llm_kwargs,
temperature=temperature, baseline_llm_kwargs,
seeded=True, test_llm_kwargs,
force_output_len=True) batch_size,
max_output_len=output_len,
temperature=temperature,
disable_seed=False,
)
# Ensure this same test does fail if we _don't_ include per-request seeds # Ensure this same test does fail if we _don't_ include per-request seeds
with pytest.raises(AssertionError): with pytest.raises(AssertionError):
run_equality_correctness_test(baseline_llm_generator, run_equality_correctness_test(
test_llm_generator, vllm_runner,
batch_size, common_llm_kwargs,
max_output_len=output_len, per_test_common_llm_kwargs,
temperature=temperature, baseline_llm_kwargs,
seeded=False, test_llm_kwargs,
force_output_len=True) batch_size,
max_output_len=output_len,
temperature=temperature,
disable_seed=True,
)