cpu_worker.py 18.1 KB
Newer Older
1
# SPDX-License-Identifier: Apache-2.0
2
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
3
"""A CPU worker class."""
4
import os
5
from importlib import util
6
from typing import List, Optional, Set, Tuple, Type
7
8
9
10

import torch
import torch.distributed

11
import vllm.envs as envs
12
from vllm.attention import get_attn_backend
13
14
from vllm.config import (CacheConfig, DeviceConfig, ModelConfig,
                         ParallelConfig, VllmConfig)
15
from vllm.distributed import (ensure_model_parallel_initialized,
16
                              init_distributed_environment)
17
from vllm.logger import init_logger
18
from vllm.lora.request import LoRARequest
19
from vllm.model_executor import set_random_seed
20
from vllm.sequence import ExecuteModelRequest
21
from vllm.utils import bind_kv_cache
22
from vllm.worker.cpu_enc_dec_model_runner import CPUEncoderDecoderModelRunner
23
from vllm.worker.cpu_model_runner import CPUModelRunner, CPUModelRunnerBase
24
from vllm.worker.cpu_pooling_model_runner import CPUPoolingModelRunner
25
from vllm.worker.worker_base import (LocalOrDistributedWorkerBase, WorkerBase,
26
                                     WorkerInput)
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56

logger = init_logger(__name__)


class CPUCacheEngine:
    """Manages the KV cache for CPU backend.

    This class is responsible for initializing and managing CPU KV
    caches. It also provides methods for performing KV cache operations, such
    as copying.
    """

    def __init__(self, cache_config: CacheConfig, model_config: ModelConfig,
                 parallel_config: ParallelConfig,
                 device_config: DeviceConfig) -> None:
        assert device_config.device_type == "cpu"
        self.cache_config = cache_config
        self.model_config = model_config
        self.parallel_config = parallel_config

        self.head_size = model_config.get_head_size()
        self.num_layers = model_config.get_num_layers(parallel_config)
        self.num_heads = model_config.get_num_kv_heads(parallel_config)

        self.block_size = cache_config.block_size
        # Note: In CacheConfig, num_gpu_blocks actual is num_cpu_blocks
        # for CPU backend, because we want to reuse KV cache management
        # in the scheduler.
        self.num_cpu_blocks = cache_config.num_gpu_blocks

57
58
        self.dtype = CPUCacheEngine.get_kv_cache_dtype(cache_config,
                                                       model_config)
59
60

        # Get attention backend.
61
62
63
64
65
        self.attn_backend = get_attn_backend(
            self.model_config.get_head_size(),
            self.model_config.dtype,
            cache_config.cache_dtype,
            self.block_size,
66
            self.model_config.is_attention_free,
Thien Tran's avatar
Thien Tran committed
67
            use_mla=self.model_config.use_mla,
68
        )
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85

        # Initialize the cache.
        self.cpu_cache = self._allocate_kv_cache(self.num_cpu_blocks)

    def _allocate_kv_cache(
        self,
        num_blocks: int,
    ) -> List[torch.Tensor]:
        """Allocates KV cache on CPU."""
        kv_cache_shape = self.attn_backend.get_kv_cache_shape(
            num_blocks, self.block_size, self.num_heads, self.head_size)
        kv_cache: List[torch.Tensor] = []
        for _ in range(self.num_layers):
            kv_cache.append(
                torch.empty(kv_cache_shape, dtype=self.dtype, device="cpu"))
        return kv_cache

86
    def swap_in(self, src_to_dst: torch.Tensor) -> None:
87
88
        raise NotImplementedError("Swap is not supported in CPUCacheEngine.")

89
    def swap_out(self, src_to_dst: torch.Tensor) -> None:
90
91
        raise NotImplementedError("Swap is not supported in CPUCacheEngine.")

92
    def copy(self, src_to_dsts: torch.Tensor) -> None:
93
94
        self.attn_backend.copy_blocks(self.cpu_cache, src_to_dsts)

95
96
97
98
99
100
101
102
103
104
105
    @staticmethod
    def get_kv_cache_dtype(cache_config: CacheConfig,
                           model_config: ModelConfig):
        if cache_config.cache_dtype == "auto":
            return model_config.dtype
        elif cache_config.cache_dtype in ["fp8", "fp8_e5m2"]:
            return torch.float8_e5m2
        else:
            raise NotImplementedError(f"Unsupported KV cache type "
                                      f"{cache_config.cache_dtype}.")

106
107
    @staticmethod
    def get_cache_block_size(
108
        cache_config: CacheConfig,
109
110
111
112
113
114
115
        model_config: ModelConfig,
        parallel_config: ParallelConfig,
    ) -> int:
        head_size = model_config.get_head_size()
        num_heads = model_config.get_num_kv_heads(parallel_config)
        num_layers = model_config.get_num_layers(parallel_config)

116
        key_cache_block = cache_config.block_size * num_heads * head_size
117
        value_cache_block = key_cache_block if not model_config.use_mla else 0
118
        total = num_layers * (key_cache_block + value_cache_block)
119
        dtype = CPUCacheEngine.get_kv_cache_dtype(cache_config, model_config)
120
121
122
123
        dtype_size = torch.tensor([], dtype=dtype).element_size()
        return dtype_size * total


124
class CPUWorker(LocalOrDistributedWorkerBase):
125
126
127
128
129
130
131
132
133
134
    """A worker class that executes (a partition of) the model on a CPU socket.

    Each worker is associated with a single CPU socket. The worker is 
    responsible for maintaining the KV cache and executing the model on the 
    CPU. In case of distributed inference, each worker is assigned a partition
    of the model.
    """

    def __init__(
        self,
135
        vllm_config: VllmConfig,
136
137
138
139
140
        local_rank: int,
        rank: int,
        distributed_init_method: str,
        kv_cache_dtype: Optional[str] = "auto",
        is_driver_worker: bool = False,
141
        model_runner_cls: Optional[Type[CPUModelRunner]] = None,
142
    ) -> None:
143
144
        WorkerBase.__init__(self, vllm_config=vllm_config)

145
146
        self.local_rank = local_rank
        self.rank = rank
147
148
        vllm_config.parallel_config.rank = rank

149
        self.distributed_init_method = distributed_init_method
150

151
152
153
        self.is_driver_worker = is_driver_worker
        if self.is_driver_worker:
            assert self.rank == 0, "The driver worker must have rank 0."
154

155
156
157
158
        if self.model_config.trust_remote_code:
            # note: lazy import to avoid importing torch before initializing
            from vllm.utils import init_cached_hf_modules
            init_cached_hf_modules()
159
160
161

        # Setup OpenMP threads affinity.
        omp_cpuids = envs.VLLM_CPU_OMP_THREADS_BIND
162
163
164
165
        self.local_omp_cpuid = "all"
        if omp_cpuids == "auto":
            self.local_omp_cpuid = self.get_cpus_id_binding_based_on_numa_nodes(
            )
166
167
168
        else:
            self.local_omp_cpuid = omp_cpuids.split("|")[rank]

169
170
171
172
173
174
175
176
177
178
        # Return hidden states from target model if the draft model is an
        # mlp_speculator
        speculative_config = self.speculative_config
        model_config = self.model_config
        speculative_args = {} if speculative_config is None \
            or (speculative_config.draft_model_config.model ==
                model_config.model) \
            or (speculative_config.draft_model_config.hf_config.model_type
                not in ["medusa", "mlp_speculator", "eagle"]) \
                    else {"return_hidden_states": True}
179
        ModelRunnerClass: Type[CPUModelRunnerBase] = CPUModelRunner
180
        if self.model_config.runner_type == "pooling":
181
            ModelRunnerClass = CPUPoolingModelRunner
182
        elif self.model_config.is_encoder_decoder:
183
            ModelRunnerClass = CPUEncoderDecoderModelRunner
184
        self.model_runner: CPUModelRunnerBase = ModelRunnerClass(
185
            vllm_config=vllm_config,
186
            kv_cache_dtype=kv_cache_dtype,
187
188
189
190
191
            is_driver_worker=is_driver_worker,
            **speculative_args,
        )
        if model_runner_cls is not None:
            self.model_runner = model_runner_cls(self.model_runner)
192
        # Uninitialized cache engine. Will be initialized by
193
        # initialize_cache.
194
        self.cache_engine: List[CPUCacheEngine]
195
        # Initialize cpu_cache as pooling models don't initialize kv_caches
196
        self.cpu_cache: Optional[List[List[torch.Tensor]]] = None
197

198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
        # Torch profiler. Enabled and configured through env vars:
        # VLLM_TORCH_PROFILER_DIR=/path/to/save/trace
        if envs.VLLM_TORCH_PROFILER_DIR:
            torch_profiler_trace_dir = envs.VLLM_TORCH_PROFILER_DIR
            logger.info("Profiling enabled. Traces will be saved to: %s",
                        torch_profiler_trace_dir)
            self.profiler = torch.profiler.profile(
                activities=[
                    torch.profiler.ProfilerActivity.CPU,
                ],
                with_stack=True,
                on_trace_ready=torch.profiler.tensorboard_trace_handler(
                    torch_profiler_trace_dir, use_gzip=True))
        else:
            self.profiler = None

    def start_profile(self):
        if self.profiler is None:
            raise RuntimeError("Profiler is not enabled.")
        self.profiler.start()

    def stop_profile(self):
        if self.profiler is None:
            raise RuntimeError("Profiler is not enabled.")
        self.profiler.stop()

224
    def init_device(self) -> None:
225
        if self.local_omp_cpuid != "all":
226
            ret = torch.ops._C_utils.init_cpu_threads_env(self.local_omp_cpuid)
227
228
            if ret:
                logger.info(ret)
229
230
231
232

        # Note: unique identifier for creating allreduce shared memory
        os.environ["VLLM_DIST_IDENT"] = self.distributed_init_method.split(
            ":")[-1]
233
        self.device = torch.device("cpu")
234
235
236
237
238
239
240
        self.init_distributed_environment()
        # Set random seed.
        set_random_seed(self.model_config.seed)

    def load_model(self):
        self.model_runner.load_model()

241
    def determine_num_available_blocks(self) -> Tuple[int, int]:
242
243
244
245
246
247
248
249
250
        """Determine the number of blocks available for the KV cache.

        This determines how many KV blocks can fit into the configured CPU
        KV cache space.

        Note that since vLLM assumes a block resides on GPU if it can be
        modified, we return num_gpu_blocks=num_cpu_blocks and num_cpu_blocks=0.
        This allows us to reuse the scheduler of vLLM without generalizing it
        to different devices.
251
252
253
        """
        # For CPU device, the block number will be calculated based on the
        # cpu_kvcache_space.
254
255
256
        cache_block_size = self.get_cache_block_size_bytes()
        num_cpu_blocks = int(self.cache_config.cpu_kvcache_space_bytes //
                             cache_block_size)
257
258
        num_cpu_blocks = max(num_cpu_blocks, 0)

259
260
261
262
263
        # Note: To reuse the cache management procedure,
        # use cpu cache as 'gpu cache'.
        num_gpu_blocks = num_cpu_blocks
        num_cpu_blocks = 0
        return num_gpu_blocks, num_cpu_blocks
264

265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
    def initialize_cache(self, num_gpu_blocks: int,
                         num_cpu_blocks: int) -> None:
        """Initialize the KV cache. Currently, swappable CPU memory is not
        supported.

        Since this worker does not support GPUs, we use the num_gpu_blocks to
        determine how many non-swappable CPU blocks to allocate.
        """
        assert (num_cpu_blocks == 0
                ), f"{type(self)} does not support swappable cache"

        # Note: To reuse the cache management procedure,
        # use cpu cache as 'gpu cache'.
        num_cpu_blocks = num_gpu_blocks

        self._validate_num_cpu_blocks(num_cpu_blocks)
        self.cache_config.num_gpu_blocks = num_cpu_blocks
        self.cache_config.num_cpu_blocks = 0

        # Initialize the cache.
        self._init_cache_engine()

287
288
289
290
291
292
293
294
295
296
297
298
    def add_lora(self, lora_request: LoRARequest) -> bool:
        return self.model_runner.add_lora(lora_request)

    def remove_lora(self, lora_id: int) -> bool:
        return self.model_runner.remove_lora(lora_id)

    def pin_lora(self, lora_id: int) -> bool:
        return self.model_runner.pin_lora(lora_id)

    def list_loras(self) -> Set[int]:
        return self.model_runner.list_loras()

299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
    def _validate_num_cpu_blocks(self, num_cpu_blocks: int) -> None:
        """Raise errors if the num_cpu_blocks is invalid.
        """
        if num_cpu_blocks <= 0:
            raise ValueError("No available memory for the cache blocks. "
                             "Try increasing `VLLM_CPU_KVCACHE_SPACE` when "
                             "initializing the engine.")

        max_seq_len = self.cache_config.block_size * num_cpu_blocks
        if self.model_config.max_model_len > max_seq_len:
            raise ValueError(
                f"The model's max seq len ({self.model_config.max_model_len}) "
                "is larger than the maximum number of tokens that can be "
                f"stored in KV cache ({max_seq_len}). Try increasing "
                "`VLLM_CPU_KVCACHE_SPACE` or decreasing `max_model_len` when "
                "initializing the engine.")

    def _init_cache_engine(self) -> None:
317
318
319
320
321
322
323
324
325
        self.cache_engine = [
            CPUCacheEngine(self.cache_config, self.model_config,
                           self.parallel_config, self.device_config)
            for _ in range(self.parallel_config.pipeline_parallel_size)
        ]
        self.cpu_cache = [
            self.cache_engine[ve].cpu_cache
            for ve in range(self.parallel_config.pipeline_parallel_size)
        ]
326
327
        bind_kv_cache(self.compilation_config.static_forward_context,
                      self.cpu_cache)
328
329
330
331
332
        self.model_runner.block_size = self.cache_engine[0].block_size

        assert all(
            self.cpu_cache[ve] is not None
            for ve in range(self.parallel_config.pipeline_parallel_size))
333
334

        # Populate the cache to warmup the memory
335
336
337
        for ve in range(self.parallel_config.pipeline_parallel_size):
            for layer_cache in self.cpu_cache[ve]:
                layer_cache.fill_(0)
338

339
340
341
342
343
    @property
    def do_metadata_broadcast(self) -> bool:
        return self.parallel_config.tensor_parallel_size > 1

    @property
344
    def kv_cache(self) -> Optional[List[List[torch.Tensor]]]:
345
346
        return self.cpu_cache

347
348
349
350
351
352
353
354
    @property
    def vocab_size(self) -> int:
        return self.model_runner.vocab_size

    @property
    def max_model_len(self) -> int:
        return self.model_config.max_model_len

355
    def execute_worker(
356
        self,
357
        worker_input: WorkerInput,
358
    ) -> None:
359
360
        if (worker_input.blocks_to_copy is not None
                and worker_input.blocks_to_copy.numel() > 0):
361
362
            self.cache_engine[worker_input.virtual_engine].copy(
                worker_input.blocks_to_copy)
363
364

    @torch.inference_mode()
365
366
367
    def prepare_worker_input(
            self, execute_model_req: ExecuteModelRequest) -> WorkerInput:
        assert execute_model_req is not None
368
        virtual_engine: int = execute_model_req.virtual_engine
369
370
371
372
373
374
375
376
377
        num_seq_groups: int = len(execute_model_req.seq_group_metadata_list)
        blocks_to_copy = torch.tensor(execute_model_req.blocks_to_copy,
                                      device="cpu",
                                      dtype=torch.int64).view(-1, 2)
        assert len(execute_model_req.blocks_to_swap_in) == 0
        assert len(execute_model_req.blocks_to_swap_out) == 0
        return WorkerInput(
            num_seq_groups=num_seq_groups,
            blocks_to_copy=blocks_to_copy,
378
            virtual_engine=virtual_engine,
379
        )
380
381
382
383
384
385
386

    def init_distributed_environment(self) -> None:
        """Initialize the distributed environment."""

        parallel_config = self.parallel_config
        rank = self.rank
        distributed_init_method = self.distributed_init_method
387
388
389
390
391
392
        init_distributed_environment(
            world_size=parallel_config.world_size,
            rank=rank,
            distributed_init_method=distributed_init_method,
            backend="gloo",
        )
393
394
395
396
397
398

        # A small all_reduce for warmup.
        torch.distributed.all_reduce(torch.zeros(1).cpu())

        ensure_model_parallel_initialized(
            parallel_config.tensor_parallel_size,
399
            parallel_config.pipeline_parallel_size)
400
401
402
403

    def get_cache_block_size_bytes(self) -> int:
        """Return the size in bytes of a single KV cache block.
        """
404
405
406
        return CPUCacheEngine.get_cache_block_size(self.cache_config,
                                                   self.model_config,
                                                   self.parallel_config)
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452

    def get_cpus_id_binding_based_on_numa_nodes(self) -> str:
        """Return CPUs id binding based on NUMA nodes.
        """
        rank_to_cpus = self.local_omp_cpuid
        # Setup OpenMP thread affinity based on NUMA nodes automatically
        world_size = self.vllm_config.parallel_config.world_size
        libnuma_found = util.find_spec("numa") is not None
        psutil_found = util.find_spec("psutil") is not None
        if libnuma_found and psutil_found:
            import psutil
            from numa import info
            cpu_count = psutil.cpu_count(logical=False)
            cpus_allow_list = psutil.Process().cpu_affinity()
            numa_size = info.get_num_configured_nodes()
            cpu_count_per_numa = cpu_count // numa_size
            num_of_reserved_cpu = min(envs.VLLM_CPU_NUM_OF_RESERVED_CPU,
                                      cpu_count_per_numa // 2)

            # check allow node_to_cpus list
            node_to_cpus = []
            for i in range(numa_size):
                node_intersect = set(
                    info.node_to_cpus(i)).intersection(cpus_allow_list)
                if bool(node_intersect):
                    node_to_cpus.append(list(node_intersect))

            if world_size > len(node_to_cpus):
                logger.error(
                    "Auto thread-binding failed due to "
                    "world size: %d is larger than "
                    "allowed NUMA nodes number: %d."
                    "Please try to bind threads manually.", world_size,
                    len(node_to_cpus))
            else:
                end = cpu_count_per_numa - num_of_reserved_cpu
                rank_to_cpus_list = node_to_cpus[self.rank][:end]
                rank_to_cpus = ','.join(str(x) for x in rank_to_cpus_list)
                logger.info("auto thread-binding list: %s", rank_to_cpus)
        else:
            logger.warning(
                "Auto thread-binding is not supported due to "
                "the lack of package numa and psutil,"
                "fallback to no thread-binding. To get better performance,"
                "please try to manually bind threads.")
        return rank_to_cpus