[CI] Try introducing isort. (#3495)
This commit is contained in:
@@ -1,12 +1,12 @@
|
||||
from dataclasses import dataclass
|
||||
import os
|
||||
import pathlib
|
||||
from dataclasses import dataclass
|
||||
|
||||
import pytest
|
||||
|
||||
from vllm.transformers_utils.tokenizer import get_tokenizer
|
||||
from vllm.entrypoints.openai.serving_chat import OpenAIServingChat
|
||||
from vllm.entrypoints.openai.protocol import ChatCompletionRequest
|
||||
from vllm.entrypoints.openai.serving_chat import OpenAIServingChat
|
||||
from vllm.transformers_utils.tokenizer import get_tokenizer
|
||||
|
||||
chatml_jinja_path = pathlib.Path(os.path.dirname(os.path.abspath(
|
||||
__file__))).parent.parent / "examples/template_chatml.jinja"
|
||||
|
||||
+1
-1
@@ -6,8 +6,8 @@ import torch
|
||||
from transformers import AutoModelForCausalLM
|
||||
|
||||
from vllm import LLM, SamplingParams
|
||||
from vllm.transformers_utils.tokenizer import get_tokenizer
|
||||
from vllm.config import TokenizerPoolConfig
|
||||
from vllm.transformers_utils.tokenizer import get_tokenizer
|
||||
|
||||
_TEST_DIR = os.path.dirname(__file__)
|
||||
_TEST_PROMPTS = [os.path.join(_TEST_DIR, "prompts", "example.txt")]
|
||||
|
||||
@@ -1,13 +1,14 @@
|
||||
import pytest
|
||||
import time
|
||||
from typing import List
|
||||
|
||||
import pytest
|
||||
|
||||
from vllm import SamplingParams
|
||||
from vllm.block import PhysicalTokenBlock
|
||||
from vllm.core.block_manager import (UncachedBlockAllocator, BlockSpaceManager,
|
||||
AllocStatus)
|
||||
from vllm.core.block_manager import (AllocStatus, BlockSpaceManager,
|
||||
UncachedBlockAllocator)
|
||||
from vllm.sequence import Logprob, Sequence, SequenceGroup, SequenceStatus
|
||||
from vllm.utils import Device
|
||||
from vllm.sequence import Sequence, SequenceGroup, SequenceStatus, Logprob
|
||||
|
||||
from .utils import create_dummy_prompt
|
||||
|
||||
|
||||
@@ -1,10 +1,11 @@
|
||||
from typing import List
|
||||
import pytest # noqa
|
||||
import time
|
||||
from typing import List
|
||||
|
||||
import pytest # noqa
|
||||
|
||||
from vllm.config import CacheConfig, SchedulerConfig
|
||||
from vllm.core.scheduler import Scheduler
|
||||
from vllm.sequence import SequenceGroup, Logprob
|
||||
from vllm.sequence import Logprob, SequenceGroup
|
||||
|
||||
from .utils import create_dummy_prompt
|
||||
|
||||
|
||||
@@ -3,14 +3,12 @@
|
||||
Run `pytest tests/distributed/test_comm_ops.py --forked`.
|
||||
"""
|
||||
import pytest
|
||||
import torch
|
||||
import ray
|
||||
import torch
|
||||
|
||||
from vllm.model_executor.parallel_utils.communication_op import (
|
||||
tensor_model_parallel_all_reduce,
|
||||
tensor_model_parallel_all_gather,
|
||||
broadcast_tensor_dict,
|
||||
)
|
||||
broadcast_tensor_dict, tensor_model_parallel_all_gather,
|
||||
tensor_model_parallel_all_reduce)
|
||||
from vllm.test_utils import (init_test_distributed_environment,
|
||||
multi_process_tensor_parallel)
|
||||
|
||||
|
||||
@@ -1,6 +1,6 @@
|
||||
import os
|
||||
import random
|
||||
|
||||
import os
|
||||
import pytest
|
||||
import ray
|
||||
import torch
|
||||
|
||||
@@ -1,11 +1,11 @@
|
||||
# This unit test should be moved to a new
|
||||
# tests/test_guided_decoding directory.
|
||||
|
||||
from transformers import AutoTokenizer
|
||||
import torch
|
||||
from transformers import AutoTokenizer
|
||||
|
||||
from vllm.model_executor.guided_logits_processors import (RegexLogitsProcessor,
|
||||
JSONLogitsProcessor)
|
||||
from vllm.model_executor.guided_logits_processors import (JSONLogitsProcessor,
|
||||
RegexLogitsProcessor)
|
||||
|
||||
TEST_SCHEMA = {
|
||||
"type": "object",
|
||||
|
||||
@@ -1,22 +1,21 @@
|
||||
# imports for guided decoding tests
|
||||
import json
|
||||
import os
|
||||
import re
|
||||
import subprocess
|
||||
import sys
|
||||
import time
|
||||
|
||||
import sys
|
||||
import jsonschema
|
||||
import openai # use the official client for correctness check
|
||||
import pytest
|
||||
import requests
|
||||
# using Ray for overall ease of process management, parallel requests,
|
||||
# and debugging.
|
||||
import ray
|
||||
import openai # use the official client for correctness check
|
||||
import requests
|
||||
# downloading lora to test lora requests
|
||||
from huggingface_hub import snapshot_download
|
||||
|
||||
# imports for guided decoding tests
|
||||
import json
|
||||
import jsonschema
|
||||
import re
|
||||
|
||||
from vllm.transformers_utils.tokenizer import get_tokenizer
|
||||
|
||||
MAX_SERVER_START_WAIT_S = 600 # wait for server to start for 60 seconds
|
||||
|
||||
@@ -1,4 +1,5 @@
|
||||
import pytest
|
||||
|
||||
from vllm.utils import create_kv_caches_with_random
|
||||
|
||||
|
||||
|
||||
@@ -2,10 +2,10 @@ from typing import Type
|
||||
|
||||
import pytest
|
||||
import torch
|
||||
from allclose_default import get_default_atol, get_default_rtol
|
||||
|
||||
from vllm.model_executor.layers.activation import (FastGELU, GeluAndMul,
|
||||
NewGELU, SiluAndMul)
|
||||
from allclose_default import get_default_atol, get_default_rtol
|
||||
|
||||
DTYPES = [torch.half, torch.bfloat16, torch.float]
|
||||
NUM_TOKENS = [7, 83, 2048] # Arbitrary values for testing
|
||||
|
||||
@@ -3,13 +3,12 @@ from typing import List, Optional, Tuple
|
||||
|
||||
import pytest
|
||||
import torch
|
||||
from allclose_default import get_default_atol, get_default_rtol
|
||||
from xformers import ops as xops
|
||||
from xformers.ops.fmha.attn_bias import BlockDiagonalCausalMask
|
||||
|
||||
from vllm._C import ops, cache_ops
|
||||
from vllm.utils import get_max_shared_memory_bytes
|
||||
from vllm.utils import is_hip
|
||||
from allclose_default import get_default_atol, get_default_rtol
|
||||
from vllm._C import cache_ops, ops
|
||||
from vllm.utils import get_max_shared_memory_bytes, is_hip
|
||||
|
||||
FLOAT32_BYTES = torch.finfo(torch.float).bits // 8
|
||||
# This will change depending on the compute capability.
|
||||
|
||||
@@ -1,10 +1,9 @@
|
||||
import random
|
||||
from typing import Tuple
|
||||
|
||||
import pytest
|
||||
import torch
|
||||
|
||||
from typing import Tuple
|
||||
|
||||
from vllm._C import cache_ops
|
||||
|
||||
COPYING_DIRECTION = [('cuda', 'cpu'), ('cuda', 'cuda'), ('cpu', 'cuda')]
|
||||
|
||||
@@ -7,8 +7,8 @@ import torch
|
||||
from transformers import MixtralConfig
|
||||
from transformers.models.mixtral.modeling_mixtral import MixtralSparseMoeBlock
|
||||
|
||||
from vllm.model_executor.layers.fused_moe import fused_moe
|
||||
from vllm.model_executor.layers.activation import SiluAndMul
|
||||
from vllm.model_executor.layers.fused_moe import fused_moe
|
||||
from vllm.model_executor.models.mixtral import MixtralMoE
|
||||
|
||||
|
||||
|
||||
@@ -1,9 +1,10 @@
|
||||
from itertools import accumulate
|
||||
from typing import List, Optional
|
||||
|
||||
import pytest
|
||||
import torch
|
||||
from allclose_default import get_default_atol, get_default_rtol
|
||||
from itertools import accumulate
|
||||
|
||||
from vllm.model_executor.layers.rotary_embedding import get_rope
|
||||
|
||||
IS_NEOX_STYLE = [True, False]
|
||||
|
||||
@@ -1,12 +1,13 @@
|
||||
import random
|
||||
import pytest
|
||||
import time
|
||||
|
||||
import pytest
|
||||
import torch
|
||||
from vllm.attention.ops.prefix_prefill import context_attention_fwd
|
||||
from xformers import ops as xops
|
||||
from xformers.ops.fmha.attn_bias import BlockDiagonalCausalFromBottomRightMask
|
||||
|
||||
from vllm.attention.ops.prefix_prefill import context_attention_fwd
|
||||
|
||||
NUM_HEADS = [64]
|
||||
NUM_QUERIES_PER_KV = [1, 8, 64]
|
||||
HEAD_SIZES = [128]
|
||||
|
||||
@@ -1,7 +1,8 @@
|
||||
import torch
|
||||
import pytest
|
||||
import random
|
||||
|
||||
import pytest
|
||||
import torch
|
||||
|
||||
from vllm.model_executor.layers.ops.rand import seeded_uniform
|
||||
from vllm.model_executor.utils import set_random_seed
|
||||
|
||||
|
||||
@@ -1,15 +1,15 @@
|
||||
import gc
|
||||
|
||||
import torch
|
||||
import pytest
|
||||
import torch
|
||||
import triton
|
||||
import triton.language as tl
|
||||
|
||||
from vllm.model_executor.layers.ops.sample import (
|
||||
_uniform_to_exponential, sample, get_num_triton_sampler_splits,
|
||||
MAX_TRITON_N_COLS)
|
||||
from vllm.model_executor.utils import set_random_seed
|
||||
MAX_TRITON_N_COLS, _uniform_to_exponential, get_num_triton_sampler_splits,
|
||||
sample)
|
||||
from vllm.model_executor.sampling_metadata import SamplingTensors
|
||||
from vllm.model_executor.utils import set_random_seed
|
||||
|
||||
SINGLE_SPLIT_VOCAB_SIZE = 32000 # llama/mistral/mixtral vocab size
|
||||
MULTI_SPLIT_VOCAB_SIZE = MAX_TRITON_N_COLS + 100
|
||||
|
||||
@@ -2,7 +2,7 @@ import contextlib
|
||||
import gc
|
||||
import tempfile
|
||||
from collections import OrderedDict
|
||||
from unittest.mock import patch, MagicMock
|
||||
from unittest.mock import MagicMock, patch
|
||||
|
||||
import pytest
|
||||
import ray
|
||||
@@ -12,13 +12,13 @@ from huggingface_hub import snapshot_download
|
||||
|
||||
import vllm
|
||||
from vllm.config import LoRAConfig
|
||||
from vllm.model_executor.layers.sampler import Sampler
|
||||
from vllm.model_executor.layers.logits_processor import LogitsProcessor
|
||||
from vllm.model_executor.model_loader import get_model
|
||||
from vllm.model_executor.layers.linear import (ColumnParallelLinear,
|
||||
MergedColumnParallelLinear,
|
||||
RowParallelLinear)
|
||||
from vllm.model_executor.layers.logits_processor import LogitsProcessor
|
||||
from vllm.model_executor.layers.sampler import Sampler
|
||||
from vllm.model_executor.layers.vocab_parallel_embedding import ParallelLMHead
|
||||
from vllm.model_executor.model_loader import get_model
|
||||
from vllm.model_executor.parallel_utils.parallel_state import (
|
||||
destroy_model_parallel, initialize_model_parallel)
|
||||
|
||||
|
||||
@@ -1,12 +1,14 @@
|
||||
import tempfile
|
||||
from random import sample
|
||||
from typing import List, Optional
|
||||
|
||||
import peft
|
||||
import pytest
|
||||
from random import sample
|
||||
import tempfile
|
||||
from transformers import AutoModelForCausalLM
|
||||
|
||||
import vllm
|
||||
from vllm.lora.request import LoRARequest
|
||||
|
||||
from .conftest import cleanup
|
||||
|
||||
MODEL_PATH = "Felladrin/Llama-68M-Chat-v1"
|
||||
|
||||
+14
-18
@@ -1,32 +1,28 @@
|
||||
import pytest
|
||||
import random
|
||||
from copy import deepcopy
|
||||
from dataclasses import dataclass
|
||||
from typing import List, Optional, Dict, Tuple
|
||||
from typing import Dict, List, Optional, Tuple
|
||||
|
||||
import pytest
|
||||
import torch
|
||||
import torch.nn.functional as F
|
||||
|
||||
from vllm.lora.layers import (
|
||||
ColumnParallelLinearWithLoRA,
|
||||
MergedColumnParallelLinearWithLoRA,
|
||||
QKVParallelLinearWithLora,
|
||||
VocabParallelEmbeddingWithLoRA,
|
||||
RowParallelLinearWithLoRA,
|
||||
LogitsProcessorWithLoRA,
|
||||
LoRAMapping,
|
||||
BaseLayerWithLoRA,
|
||||
)
|
||||
from vllm.lora.models import (LoRALayerWeights, convert_mapping,
|
||||
PackedLoRALayerWeights)
|
||||
from vllm.config import LoRAConfig
|
||||
from vllm.model_executor.layers.logits_processor import LogitsProcessor
|
||||
from vllm.lora.layers import (BaseLayerWithLoRA, ColumnParallelLinearWithLoRA,
|
||||
LogitsProcessorWithLoRA, LoRAMapping,
|
||||
MergedColumnParallelLinearWithLoRA,
|
||||
QKVParallelLinearWithLora,
|
||||
RowParallelLinearWithLoRA,
|
||||
VocabParallelEmbeddingWithLoRA)
|
||||
from vllm.lora.models import (LoRALayerWeights, PackedLoRALayerWeights,
|
||||
convert_mapping)
|
||||
from vllm.model_executor.layers.linear import (ColumnParallelLinear,
|
||||
MergedColumnParallelLinear,
|
||||
RowParallelLinear,
|
||||
QKVParallelLinear)
|
||||
QKVParallelLinear,
|
||||
RowParallelLinear)
|
||||
from vllm.model_executor.layers.logits_processor import LogitsProcessor
|
||||
from vllm.model_executor.layers.vocab_parallel_embedding import (
|
||||
VocabParallelEmbedding, ParallelLMHead)
|
||||
ParallelLMHead, VocabParallelEmbedding)
|
||||
from vllm.model_executor.utils import set_random_seed
|
||||
|
||||
from .utils import DummyLoRAManager
|
||||
|
||||
@@ -3,6 +3,7 @@ import ray
|
||||
|
||||
import vllm
|
||||
from vllm.lora.request import LoRARequest
|
||||
|
||||
from .conftest import cleanup
|
||||
|
||||
MODEL_PATH = "meta-llama/Llama-2-7b-hf"
|
||||
|
||||
@@ -8,11 +8,11 @@ from torch import nn
|
||||
|
||||
from vllm.config import LoRAConfig
|
||||
from vllm.lora.layers import (ColumnParallelLinearWithLoRA,
|
||||
RowParallelLinearWithLoRA,
|
||||
MergedColumnParallelLinearWithLoRA)
|
||||
MergedColumnParallelLinearWithLoRA,
|
||||
RowParallelLinearWithLoRA)
|
||||
from vllm.lora.lora import LoRALayerWeights, PackedLoRALayerWeights
|
||||
from vllm.lora.models import (LoRAModel, LoRAModelManager,
|
||||
LRUCacheLoRAModelManager, LoRAMapping)
|
||||
from vllm.lora.models import (LoRAMapping, LoRAModel, LoRAModelManager,
|
||||
LRUCacheLoRAModelManager)
|
||||
from vllm.lora.request import LoRARequest
|
||||
from vllm.lora.worker_manager import (LRUCacheWorkerLoRAManager,
|
||||
WorkerLoRAManager)
|
||||
|
||||
@@ -1,8 +1,10 @@
|
||||
import pytest
|
||||
from transformers import AutoTokenizer, PreTrainedTokenizerBase
|
||||
|
||||
from vllm.lora.request import LoRARequest
|
||||
from vllm.transformers_utils.tokenizer_group import get_tokenizer_group
|
||||
from vllm.transformers_utils.tokenizer import get_lora_tokenizer
|
||||
from vllm.transformers_utils.tokenizer_group import get_tokenizer_group
|
||||
|
||||
from ..conftest import get_tokenizer_pool_config
|
||||
|
||||
|
||||
|
||||
@@ -2,8 +2,8 @@ from collections import OrderedDict
|
||||
|
||||
from torch import nn
|
||||
|
||||
from vllm.lora.utils import parse_fine_tuned_lora_name, replace_submodule
|
||||
from vllm.utils import LRUCache
|
||||
from vllm.lora.utils import (parse_fine_tuned_lora_name, replace_submodule)
|
||||
|
||||
|
||||
def test_parse_fine_tuned_lora_name():
|
||||
|
||||
@@ -3,10 +3,10 @@ import random
|
||||
import tempfile
|
||||
from unittest.mock import patch
|
||||
|
||||
from vllm.config import (DeviceConfig, LoRAConfig, ModelConfig, ParallelConfig,
|
||||
SchedulerConfig)
|
||||
from vllm.lora.models import LoRAMapping
|
||||
from vllm.lora.request import LoRARequest
|
||||
from vllm.config import (ModelConfig, ParallelConfig, SchedulerConfig,
|
||||
DeviceConfig, LoRAConfig)
|
||||
from vllm.worker.worker import Worker
|
||||
|
||||
|
||||
|
||||
@@ -11,9 +11,11 @@ up to 3 times to see if we pass.
|
||||
Run `pytest tests/models/test_marlin.py --forked`.
|
||||
"""
|
||||
|
||||
from dataclasses import dataclass
|
||||
|
||||
import pytest
|
||||
import torch
|
||||
from dataclasses import dataclass
|
||||
|
||||
from vllm.model_executor.layers.quantization import (
|
||||
_QUANTIZATION_CONFIG_REGISTRY)
|
||||
|
||||
|
||||
@@ -1,7 +1,7 @@
|
||||
import pytest
|
||||
import torch
|
||||
from tests.conftest import VllmRunner
|
||||
|
||||
from tests.conftest import VllmRunner
|
||||
from vllm import SamplingParams
|
||||
|
||||
MODELS = ["facebook/opt-125m"]
|
||||
|
||||
@@ -1,13 +1,12 @@
|
||||
"""Tests for rejection sampling."""
|
||||
import pytest
|
||||
from typing import List, Tuple
|
||||
|
||||
import pytest
|
||||
import torch
|
||||
import torch.nn.functional as F
|
||||
|
||||
from vllm.model_executor.utils import set_random_seed
|
||||
|
||||
from vllm.model_executor.layers.rejection_sampler import RejectionSampler
|
||||
from vllm.model_executor.utils import set_random_seed
|
||||
|
||||
CUDA_DEVICES = [
|
||||
f"cuda:{i}" for i in range(1 if torch.cuda.device_count() == 1 else 2)
|
||||
|
||||
@@ -1,11 +1,10 @@
|
||||
import random
|
||||
from typing import Tuple, List
|
||||
from typing import List, Optional, Tuple
|
||||
from unittest.mock import patch
|
||||
|
||||
import pytest
|
||||
import torch
|
||||
from transformers import GenerationConfig, GenerationMixin
|
||||
from typing import Optional
|
||||
|
||||
from vllm.model_executor.layers.sampler import Sampler
|
||||
from vllm.model_executor.utils import set_random_seed
|
||||
|
||||
@@ -8,8 +8,8 @@ from itertools import combinations
|
||||
|
||||
import pytest
|
||||
|
||||
from vllm.model_executor.utils import set_random_seed
|
||||
from vllm import SamplingParams
|
||||
from vllm.model_executor.utils import set_random_seed
|
||||
|
||||
MODEL = "facebook/opt-125m"
|
||||
RANDOM_SEEDS = list(range(5))
|
||||
|
||||
@@ -1,9 +1,9 @@
|
||||
import torch
|
||||
import pytest
|
||||
import torch
|
||||
|
||||
from vllm.spec_decode.batch_expansion import BatchExpansionTop1Scorer
|
||||
|
||||
from .utils import mock_worker, create_seq_group_metadata_from_prompts
|
||||
from .utils import create_seq_group_metadata_from_prompts, mock_worker
|
||||
|
||||
|
||||
@pytest.mark.parametrize('num_target_seq_ids', [100])
|
||||
|
||||
@@ -1,9 +1,9 @@
|
||||
import torch
|
||||
import math
|
||||
import pytest
|
||||
|
||||
from unittest.mock import MagicMock
|
||||
|
||||
import pytest
|
||||
import torch
|
||||
|
||||
from vllm.spec_decode.metrics import AsyncMetricsCollector
|
||||
|
||||
|
||||
|
||||
@@ -1,18 +1,19 @@
|
||||
import torch
|
||||
import random
|
||||
import pytest
|
||||
from unittest.mock import MagicMock
|
||||
|
||||
from vllm.spec_decode.multi_step_worker import (MultiStepWorker,
|
||||
DraftModelTop1Proposer)
|
||||
from vllm.worker.worker import Worker
|
||||
import pytest
|
||||
import torch
|
||||
|
||||
from vllm.model_executor.utils import set_random_seed
|
||||
from vllm.sequence import SamplerOutput
|
||||
from vllm.spec_decode.multi_step_worker import (DraftModelTop1Proposer,
|
||||
MultiStepWorker)
|
||||
from vllm.worker.worker import Worker
|
||||
|
||||
from .utils import (create_execute_model_data, create_worker,
|
||||
create_seq_group_metadata_from_prompts, zero_kv_cache,
|
||||
patch_execute_model_with_seeds,
|
||||
assert_logprobs_dict_allclose, create_batch)
|
||||
from .utils import (assert_logprobs_dict_allclose, create_batch,
|
||||
create_execute_model_data,
|
||||
create_seq_group_metadata_from_prompts, create_worker,
|
||||
patch_execute_model_with_seeds, zero_kv_cache)
|
||||
|
||||
|
||||
@pytest.mark.parametrize('num_steps', list(range(1, 17)))
|
||||
|
||||
@@ -1,18 +1,20 @@
|
||||
import torch
|
||||
import random
|
||||
import pytest
|
||||
from unittest.mock import MagicMock
|
||||
|
||||
import pytest
|
||||
import torch
|
||||
|
||||
from vllm.model_executor.layers.rejection_sampler import RejectionSampler
|
||||
from vllm.model_executor.utils import set_random_seed
|
||||
from vllm.spec_decode.interfaces import SpeculativeProposals
|
||||
from vllm.spec_decode.metrics import (AsyncMetricsCollector,
|
||||
SpecDecodeWorkerMetrics)
|
||||
from vllm.spec_decode.multi_step_worker import MultiStepWorker
|
||||
from vllm.spec_decode.spec_decode_worker import (SpecDecodeWorker,
|
||||
split_num_cache_blocks_evenly)
|
||||
from vllm.spec_decode.interfaces import SpeculativeProposals
|
||||
from vllm.model_executor.utils import set_random_seed
|
||||
from vllm.model_executor.layers.rejection_sampler import RejectionSampler
|
||||
from .utils import (mock_worker, create_batch, ExecuteModelData,
|
||||
create_sampler_output_list)
|
||||
from vllm.spec_decode.metrics import (SpecDecodeWorkerMetrics,
|
||||
AsyncMetricsCollector)
|
||||
|
||||
from .utils import (ExecuteModelData, create_batch, create_sampler_output_list,
|
||||
mock_worker)
|
||||
|
||||
|
||||
@pytest.mark.parametrize('k', [1, 2, 6])
|
||||
|
||||
@@ -1,9 +1,9 @@
|
||||
from vllm.spec_decode.util import get_all_seq_ids
|
||||
from vllm.sequence import SequenceGroupMetadata
|
||||
from vllm.spec_decode.util import split_batch_by_proposal_len
|
||||
from unittest.mock import MagicMock
|
||||
|
||||
import pytest
|
||||
from unittest.mock import MagicMock
|
||||
|
||||
from vllm.sequence import SequenceGroupMetadata
|
||||
from vllm.spec_decode.util import get_all_seq_ids, split_batch_by_proposal_len
|
||||
|
||||
|
||||
def test_get_all_seq_ids():
|
||||
|
||||
+12
-10
@@ -1,17 +1,19 @@
|
||||
import torch
|
||||
from typing import List, Optional, Dict, Iterable, Union
|
||||
from dataclasses import dataclass, fields
|
||||
from itertools import count
|
||||
from typing import Dict, Iterable, List, Optional, Union
|
||||
from unittest.mock import MagicMock
|
||||
|
||||
from vllm.worker.worker import Worker
|
||||
from vllm.utils import get_distributed_init_method, get_ip, get_open_port
|
||||
import torch
|
||||
|
||||
from vllm.engine.arg_utils import EngineArgs
|
||||
from vllm.sequence import (Logprob, SequenceGroupMetadata, SequenceData,
|
||||
SamplerOutput, SequenceGroupOutput, SequenceOutput)
|
||||
from vllm.sampling_params import SamplingParams
|
||||
from vllm.worker.cache_engine import CacheEngine
|
||||
from vllm.model_executor.utils import set_random_seed
|
||||
from itertools import count
|
||||
from dataclasses import dataclass, fields
|
||||
from vllm.sampling_params import SamplingParams
|
||||
from vllm.sequence import (Logprob, SamplerOutput, SequenceData,
|
||||
SequenceGroupMetadata, SequenceGroupOutput,
|
||||
SequenceOutput)
|
||||
from vllm.utils import get_distributed_init_method, get_ip, get_open_port
|
||||
from vllm.worker.cache_engine import CacheEngine
|
||||
from vllm.worker.worker import Worker
|
||||
|
||||
|
||||
@dataclass
|
||||
|
||||
@@ -7,8 +7,8 @@ from typing import List, Optional
|
||||
import pytest
|
||||
|
||||
from vllm.lora.request import LoRARequest
|
||||
from vllm.transformers_utils.tokenizer_group import TokenizerGroup
|
||||
from vllm.sequence import Sequence
|
||||
from vllm.transformers_utils.tokenizer_group import TokenizerGroup
|
||||
|
||||
# Make two prefixes with different first blocks.
|
||||
prefix_start = [("You are an expert"), ("You are a")]
|
||||
|
||||
@@ -1,6 +1,6 @@
|
||||
import pytest
|
||||
|
||||
from vllm.sequence import SequenceGroupOutput, SamplerOutput, SequenceOutput
|
||||
from vllm.sequence import SamplerOutput, SequenceGroupOutput, SequenceOutput
|
||||
|
||||
|
||||
@pytest.fixture
|
||||
|
||||
@@ -1,7 +1,9 @@
|
||||
from copy import deepcopy
|
||||
from vllm.transformers_utils.tokenizer import get_cached_tokenizer
|
||||
|
||||
from transformers import AutoTokenizer
|
||||
|
||||
from vllm.transformers_utils.tokenizer import get_cached_tokenizer
|
||||
|
||||
|
||||
def test_cached_tokenizer():
|
||||
reference_tokenizer = AutoTokenizer.from_pretrained("gpt2")
|
||||
|
||||
@@ -1,12 +1,12 @@
|
||||
from typing import Dict, List
|
||||
|
||||
import pytest
|
||||
|
||||
from transformers import AutoTokenizer
|
||||
from typing import List, Dict
|
||||
|
||||
from vllm.sequence import Sequence, Logprob, SamplingParams, SequenceGroup
|
||||
from vllm.transformers_utils.tokenizer_group import get_tokenizer_group
|
||||
from vllm.transformers_utils.tokenizer import detokenize_incrementally
|
||||
from vllm.sequence import Logprob, SamplingParams, Sequence, SequenceGroup
|
||||
from vllm.transformers_utils.detokenizer import Detokenizer
|
||||
from vllm.transformers_utils.tokenizer import detokenize_incrementally
|
||||
from vllm.transformers_utils.tokenizer_group import get_tokenizer_group
|
||||
|
||||
TRUTH = [
|
||||
"Hello here, this is a simple test",
|
||||
|
||||
@@ -1,14 +1,16 @@
|
||||
import os
|
||||
import pytest
|
||||
import asyncio
|
||||
import os
|
||||
from unittest.mock import patch
|
||||
|
||||
import pytest
|
||||
from transformers import AutoTokenizer, PreTrainedTokenizerBase
|
||||
|
||||
from vllm.transformers_utils.tokenizer_group import get_tokenizer_group
|
||||
from vllm.transformers_utils.tokenizer_group.ray_tokenizer_group import (
|
||||
RayTokenizerGroupPool)
|
||||
from vllm.transformers_utils.tokenizer_group.tokenizer_group import (
|
||||
TokenizerGroup)
|
||||
|
||||
from ..conftest import get_tokenizer_pool_config
|
||||
|
||||
|
||||
|
||||
@@ -1,8 +1,8 @@
|
||||
import torch
|
||||
|
||||
from vllm.engine.arg_utils import EngineArgs
|
||||
from vllm.worker.worker import Worker
|
||||
from vllm.utils import get_distributed_init_method, get_ip, get_open_port
|
||||
from vllm.worker.worker import Worker
|
||||
|
||||
|
||||
def test_swap() -> None:
|
||||
|
||||
Reference in New Issue
Block a user