Files
FastDeploy/tests/cache_manager/v1/test_cache_controller.py
T
kevin 7707be8384 [Feature][KVCache] Implement Cache Manager V1 with GPU + CPU Cache Support (1/n) (#7097)
* [Feature][KVCache] Support cache manager v1 architecture

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* Update cache manager and related modules

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* chore: update cache_manager and related modules

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* fix: add node to evictable set in complete_swap_to_device

When a node transitions from SWAP_TO_DEVICE to DEVICE via
complete_swap_to_device, it was not being added to the
_evictable_device set. This caused nodes with ref_count=0 to
become "orphaned" - not appearing in any evictable set despite
having cache_status=DEVICE.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* feat: update cache manager v1 and related modules

- Add new cache_manager.py with cache management functionality
- Add radix_tree.py for prefix caching
- Update block_pool.py and metadata.py
- Update request.py and resource_manager_v1.py for scheduling
- Update gpu_model_runner.py for GPU model execution

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* feat(cache): add cache controller v1 implementation

- Add CacheController class for cache management
- Update config.py with cache related configurations
- Refactor gpu_model_runner.py for improved cache handling

* feat(cache_manager): update cache manager v1

* fix(cache_manager): 修复 swap_cache H2D/D2H 方向的 block_ids 逻辑并清理 ForwardMeta

## Motivation

修复 swap_cache_optimized.cu 中 H2D 方向时 src/dst block_ids 使用错误的问题,
并清理 ForwardMeta 中已废弃的 cache_controller 字段。

## Modifications

- fix: swap_cache_optimized.cu 中根据 D2H 模板参数正确选取 src/dst block_ids,
  修复 H2D 方向 src/dst 倒置 bug(同时修复 SwapCachePerLayerImpl 和 SwapCacheAllLayersBatchImpl)
- refactor: cache_manager/v1/__init__.py 将 LayerSwapTimeoutError 导入从
  cache_controller 改为 cache_utils(正确来源)
- refactor: ForwardMeta 移除废弃的 cache_controller 字段
- refactor: gpu_model_runner.py 移除对应的 cache_controller 赋值语句
- test: 新增 tests/cache_manager/v1/test_swap_cache_ops.py 单元测试

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

* feat(cache_manager): refactor cache manager v1 and optimize swap ops

## Motivation

对 cache manager v1 进行重构和优化,精简代码结构,提升可维护性。

## Modifications

- 重构 transfer_manager.py,大幅精简代码逻辑
- 优化 swap_cache_optimized.cu GPU 算子实现
- 调整 cache_manager.py、cache_controller.py 逻辑,修复 free_device_blocks 方法缺失问题
- 更新 block_pool.py、cache_utils.py、metadata.py、radix_tree.py
- 精简 gpu_model_runner.py、forward_meta.py、attention.py 中相关调用
- 更新对应单元测试(test_cache_controller、test_swap_cache_ops、test_transfer_manager)
- 调整 config.py 中相关配置项

* [KVCache][MTP] 支持 cache_manager_v1 下的 MTP KV Cache 初始化及多模态 hash

## Motivation

在 enable_cache_manager_v1 路径下,MTP(speculative decode)的 KV Cache 需要由
CacheController 统一管理,以复用 swap/transfer 能力,同时修复多模态场景下 block
hash 未携带 multimodal extra_keys 的问题。

## Modifications

- `cache_controller.py`
  - 新增 `initialize_mtp_kv_cache`:通过 CacheController 初始化 MTP KV Cache,
    并将其注册到 cache_kvs_map,使 transfer_manager 自动覆盖 MTP 层
  - `initialize_host_cache` 中的 num_layers 改为包含 MTP 额外 cache 层数,保证
    Host Cache 也为 MTP 分配足够空间
  - `_free_gpu_cache` 改名为 `free_gpu_cache`(对外可调用)

- `cache_utils.py`
  - 新增 `get_block_hash_extra_keys`:提取单个 block 内的多模态 hash 信息,
    对齐 PrefixCacheManager 的 multimodal extra_keys 逻辑
  - `get_request_block_hasher` 中在 hash_block_tokens 时携带 extra_keys,
    修复多模态场景 prefix cache 命中率不准的问题

- `spec_decode/mtp.py`
  - `update_mtp_block_num` 新增 `skip_cache_init` 参数,避免 v1 cache manager
    路径下重复初始化 MTP KV Cache

- `gpu_model_runner.py`
  - `initialize_kv_cache(v1)` 路径:在主模型 cache 初始化后,调用
    `cache_controller.initialize_mtp_kv_cache` 完成 MTP cache 创建
  - `clear_cache` / `wakeup` / `reset` 等路径:respect `enable_cache_manager_v1`
    标志,跳过重复的 proposer.initialize_kv_cache 调用

## Usage or Command

```bash
# 启动支持 MTP + cache_manager_v1 的推理服务(示例)
bash run.sh
```

* fix(cache_manager): multi-GPU fix, mm hash boundary fix, and remove batch ops

1. Fix CuPy stream/event creation for multi-GPU: wrap all stream operations
   with cp.cuda.Device(device_id) context to ensure streams/events are bound
   to the correct device, preventing cross-device errors in multi-GPU setups.

2. Remove cudaSetDevice from SwapCacheAllLayers (handled by cupy context now).

3. Remove swap_cache_all_layers_batch op: simplified the implementation by
   removing the batch upload variant; all-layer transfers now use the standard
   swap_cache_all_layers with cupy device context.

4. Fix mm hash boundary comparison in get_block_hash_extra_keys: change
   strict less-than (<) to less-than-or-equal (<=) so that multimodal items
   ending exactly at block start are correctly excluded.

5. Extract config fields to KVCacheBase: model_config, cache_config,
   quant_config, parallel_config are now set in the base class __init__ to
   avoid duplication in CacheController and CacheManager subclasses.

6. Translate metadata.py docstrings from Chinese to English for broader
   contributor accessibility.

7. Add test_cache_utils.py: comprehensive unit tests for
   get_block_hash_extra_keys covering all boundary and overlap scenarios.

8. Expand test suite: test_request.py cache fields tests, test_radix_tree.py
   backup candidate tests, test_transfer_manager.py and test_cache_manager.py
   multi-GPU and concurrent operation tests.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

* [BugFix][KVCache] fix List import and move write_policy normalization to CacheManager

## Motivation

修复两处问题:
1. `fastdeploy/engine/request.py` 中 `List` 未导入导致 pre-commit F821 报错
2. `write_policy` 归一化逻辑(`write_through` → `write_through_selective`)不应放在 `FDConfig`,移至 `CacheManager.__init__` 中,使其只影响 Cache Manager V1 的内部逻辑

## Modifications

- `fastdeploy/engine/request.py`: 在 `typing` 导入中补充 `List`,删除重复的 `CacheSwapMetadata` TYPE_CHECKING 导入,修复 F821/F811
- `fastdeploy/config.py`: 删除 `write_policy` 归一化逻辑
- `fastdeploy/cache_manager/v1/cache_manager.py`: 将归一化逻辑移入 `CacheManager.__init__`

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

* [BugFix][KVCache] fix pre-commit code style issues

## Motivation

修复 CI pre-commit 代码风格检查失败问题。

## Modifications

- `fastdeploy/engine/common_engine.py`: black 格式化
- `fastdeploy/worker/worker_process.py`: black 格式化 + isort 修复
- `fastdeploy/cache_manager/v1/storage/__init__.py`: isort 修复
- `fastdeploy/worker/gpu_worker.py`: isort 修复

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

* [Feature][KVCache] update cache_manager_v1 modules

## Motivation

更新 Cache Manager V1 相关模块,完善版权信息、改进模块结构与可维护性。

## Modifications

- `fastdeploy/cache_manager/v1/` 系列模块:补充版权 header,优化代码结构
- `fastdeploy/config.py`:配置项更新
- `fastdeploy/engine/sched/resource_manager_v1.py`:调度相关更新

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

* [Feature][KVCache] add BatchRequest.from_tasks and refactor worker task parsing

## Motivation

将 worker_process 中重复的 task 解析逻辑收敛到 BatchRequest,减少代码冗余,提升可维护性。

## Modifications

- `fastdeploy/engine/request.py`:新增 `BatchRequest.from_tasks()` 类方法,统一将 task_queue 任务分类为推理请求和控制请求
- `fastdeploy/worker/worker_process.py`:使用 `BatchRequest.from_tasks()` 替代内联解析逻辑,并修复重复的 control_reqs 处理块

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

* [Feature][KVCache] add NUMA affinity for host cache and skip swap cache tests

## Motivation

优化 Host cache 内存分配的 NUMA 亲和性,减少跨 NUMA 访问延迟;
同时跳过 swap cache ops 测试(当前环境不支持)。

## Modifications

- `fastdeploy/cache_manager/v1/cache_controller.py`:
  - 新增 `_get_numa_node_for_gpu()` 方法,通过 nvidia-smi 或 sysfs 获取 GPU 对应的 NUMA 节点
  - 新增 `_bind_to_closest_numa_node()` 方法,绑定当前线程到 GPU 最近的 NUMA 节点
  - 在 `initialize_host_cache()` 中调用 NUMA 绑定,优化 H2D 传输性能
- `tests/cache_manager/v1/test_swap_cache_ops.py`:跳过所有测试类(`TestSwapCacheAllLayersCorrectness`、`TestSwapCacheAllLayersPerformance`、`TestSwapCacheRandomBlockIndices`)

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

* [BugFix][KVCache] fix unittest failures for cache_manager_v1

三个单测因接口变更或 Mock 方式问题导致失败,需修复。

- tests/distributed/chunked_moe.py:`setup_model_runner` 使用 `__new__` 跳过 `__init__`,补加 `enable_cache_manager_v1 = False`,修复 `AttributeError`
- tests/engine/test_resource_manager.py:`PrefixCacheManager` 为局部导入,`patch` 路径改为定义位置 `fastdeploy.cache_manager.prefix_cache_manager.PrefixCacheManager`
- tests/v1/test_resource_manager_v1.py:`_trigger_preempt` 第四参数已由 `list` 改为 `BatchRequest`,更新测试传参和断言

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

* [BugFix][KVCache] remove debug logging code

## Modifications

- fastdeploy/engine/request.py:删除调试用 logger 及 prompt_hashes 中的 debug 日志
- fastdeploy/worker/worker_process.py:删除 __main__ 中的调试 import 和 print 语句

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

* [BugFix][KVCache] fix cupy device id caching and pickle for _match_result

## Motivation

修复两个 bug:
1. `transfer_manager.py` 中每次调用 `cp.cuda.runtime.getDevice()` 存在隐患,应在初始化时缓存为实例变量,保证后续操作使用一致的设备 ID。
2. `request.py` 的 `__getstate__` 未跳过 `_match_result`,该字段包含 BlockNode 树的父子循环引用,pickle 时会触发 `RecursionError`;同时补充 `__setstate__` 确保 unpickle 后字段恢复为安全默认值。

## Modifications

- `transfer_manager.py`:初始化时调用 `cp.cuda.runtime.getDevice()` 并缓存到 `self._cupy_device_id`,后续 `with cp.cuda.Device(...)` 和日志均使用该缓存值。
- `request.py`:
  - `__getstate__` 中将 `_match_result` 加入跳过集合 `_SKIP_KEYS`,避免循环引用导致 pickle 失败。
  - 新增 `__setstate__`,unpickle 后将 `_block_hasher` 和 `_match_result` 恢复为 `None`。

## Usage or Command

* fix(test): fix unit test errors for _trigger_preempt and wakeup with MTP

- Use BatchRequest instead of list in test_trigger_preempt_records_tasks
- Add missing enable_cache_manager_v1 attr in TestSleepWakeupBehavior._make_runner

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

* [BugFix][KVCache] fix gpu_free_block_list returning wrong block IDs

## Motivation

`gpu_free_block_list` 的兼容 property 中误用了 `list(range(N))`,
将 `available_blocks()` 的返回值当作整数传给 `range()`,
导致返回 `[0, 1, ..., N-1]` 的假列表,而非真实的空闲 block ID。

## Modifications

- `cache_manager/v1/cache_manager.py`:将 `list(range(self._device_pool.available_blocks()))` 改为 `list(self._device_pool.available_blocks())`

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

* [BugFix][KVCache] 修复 gpu_free_block_list 返回 int 导致 TypeError

## Motivation

gpu_free_block_list 属性中调用 BlockPool.available_blocks(),
该方法返回 int(空闲块数量),用 list() 包装 int 会触发
TypeError: 'int' object is not iterable。

## Modifications

将 list(self._device_pool.available_blocks()) 改为
list(self._device_pool._free_blocks),直接返回空闲块索引列表。

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

* [KVCache][CacheManager] 适配 V1 CacheManager 的 pause/sleep/free_cache 操作

## Motivation

V1 CacheManager 引入了新的 reset_cache() 接口,pause 和 sleep 操作需要适配,
同时 free_cache 需要支持可选的 clear_storage 参数。

## Modifications

- cache_controller.py: free_cache 新增 clear_storage 参数(默认 False),
  仅当 clear_storage=True 时才调用 _clear_storage(),避免不必要的 storage 清空
- common_engine.py: pause 和 sleep 操作中,当 ENABLE_V1_KVCACHE_MANAGER 时
  使用 cache_manager.reset_cache() 替代旧的 reset() 和 pause_transfer 逻辑
- gpu_model_runner.py: sleep 时仅在非 V1 cache manager 下执行 MTP cache 清除

## Usage or Command

# 启动服务(V1 CacheManager)
python -m fastdeploy.entrypoints.openai.api_server \
  --enable-v1-kvcache-manager \
  ...

* [BugFix][KVCache] fix missing enable_cache_manager_v1 in test mocks and remove unused select_blocks_for_backup

- Remove unused `select_blocks_for_backup` method from radix_tree.py
- Fix `match_prefix` default param `skip_storage=True` and log order in cache_manager.py
- Sync test_gpu_model_runner.py with upstream/develop (add TestInsertTasksV1SplitwiseSuffix)
- Add `enable_cache_manager_v1=False` to all mock runners to fix AttributeError in CI

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

* [BugFix][KVCache] simplify _free_blocks in ResourceManagerV1 for non-v1 path

Remove redundant prefix_caching branch in else path; always call
recycle_gpu_blocks with full block_tables for non-cache-manager-v1 case.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

* [KVCache][Optimization][BugFix] fix and optimize block_pool, cache_manager, transfer_manager, request

## Motivation

修复 cache_manager v1 中若干代码质量问题,提升性能并消除潜在的类型不一致 Bug。

## Modifications

1. **block_pool.py**:`BlockPool.allocate` 将逐个 pop 循环替换为切片 + 批量 set.update,消除 Python 循环开销,O(n) → O(k)(C 层批量操作)
2. **cache_manager.py**:`match_prefix` 在 prefix caching 关闭时提前 return 前写入空 `MatchResult()`,避免调用方解引用 `_match_result=None` 崩溃
3. **transfer_manager.py**:`_build_device_layer_indices` 在 `_cache_kvs_map` 为空时也重置四个层索引列表,防止残留旧 tensor 被 swap 算子使用
4. **request.py**:`BatchRequest.append_swap_metadata` / `append_evict_metadata` 构造 `CacheSwapMetadata` 时将 `src_type`/`dst_type` 从字符串改为 `CacheLevel` 枚举,与字段类型声明一致;补充 `CacheLevel` 导入;`match_result` 属性返回类型标注修正为 `Optional[MatchResult]`
5. **resource_manager_v1.py**:`_allocate_gpu_blocks` 日志从 `INFO` 降级为 `DEBUG`,消除高频调度路径的日志噪音
6. **tests/engine/test_request.py**:同步更新 `src_type`/`dst_type` 断言为 `CacheLevel` 枚举值,补充 `CacheLevel` 导入

## Usage or Command

单元测试:
```bash
source .venv/py310/bin/activate
cd baidu/FastDeploy
python -m pytest tests/cache_manager/v1/test_cache_manager.py -v
python -m pytest tests/cache_manager/v1/test_transfer_manager.py -v
python -m pytest tests/engine/test_request.py -v
```

* [BugFix][KVCache] Fix BlockPool.allocate returns all blocks when num_blocks=0

## Motivation

当 `allocate(num_blocks=0)` 被调用时,Python 负索引陷阱导致严重错误:
`-0 == 0`,所以 `self._free_blocks[-0:]` 等价于 `self._free_blocks[0:]`,
会返回并清空整个空闲块列表,而非返回空列表。

## Modifications

在 `BlockPool.allocate` 中增加对 `num_blocks == 0` 的提前判断,直接返回 `[]`,
避免触发 Python 负索引陷阱。

## Usage or Command

```bash
# 运行相关单元测试验证修复
python -m pytest tests/cache_manager/v1/test_cache_manager.py -vv -s
```

* [KVCache][Test] add unit tests for cache_manager v1 modules

## Motivation

补全 cache_manager/v1 各模块的单测覆盖,确保核心方法有完整的测试保障。

## Modifications

新增/补充以下测试文件,全部 326 个用例通过:

- tests/cache_manager/v1/test_block_pool.py(新建)
  覆盖 BlockPool.get_metadata/set_metadata/resize、DeviceBlockPool/HostBlockPool
- tests/cache_manager/v1/test_metadata.py(新建)
  覆盖 BlockNode、RadixTreeStats、MatchResult、CacheSwapMetadata、AsyncTaskHandler
- tests/cache_manager/v1/test_cache_utils.py(补充)
  新增 hash_block_tokens、get_request_block_hasher、LayerDoneCounter 时间追踪及内部辅助方法
- tests/cache_manager/v1/test_radix_tree.py(补充)
  新增 TestCompleteSwapToDevice 专项测试类(6 个用例)
- tests/cache_manager/v1/test_cache_manager.py(补充)
  新增 offload_to_host、load_from_host、pending backup 系列、prepare_prefetch_metadata
- tests/cache_manager/v1/test_transfer_manager.py(补充)
  新增 _swap_single_layer 校验路径、sync_input/output_stream、record_input_stream_event

## Usage or Command

```bash
# 运行所有新增单测
source .venv/py310/bin/activate
python -m pytest tests/cache_manager/v1/test_block_pool.py \
  tests/cache_manager/v1/test_metadata.py \
  tests/cache_manager/v1/test_cache_utils.py \
  tests/cache_manager/v1/test_radix_tree.py \
  tests/cache_manager/v1/test_cache_manager.py \
  tests/cache_manager/v1/test_transfer_manager.py -v
# 期望结果:326 passed
```

---------

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
Co-authored-by: Jiang-Jia-Jun <163579578+Jiang-Jia-Jun@users.noreply.github.com>
2026-04-21 14:39:00 +08:00

728 lines
27 KiB
Python

# Copyright (c) 2025 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""
Unit tests for CacheController class with the new LayerDoneCounter design.
Tests cover:
- Initialization
- load_host_to_device returns LayerDoneCounter
- evict_device_to_host returns LayerDoneCounter
- submit_swap_tasks returns LayerDoneCounter
- LayerDoneCounter methods: wait_for_layer, wait_all, mark_layer_done, mark_all_done
- Statistics
- Edge cases (empty metadata, failed transfers)
"""
import time
import unittest
from unittest.mock import MagicMock, patch
from utils import get_default_test_fd_config
from fastdeploy.cache_manager.v1.metadata import CacheSwapMetadata
def create_cache_controller(
enable_prefix_caching: bool = True,
num_host_blocks: int = 50,
num_layers: int = 4,
):
"""Helper to create CacheController with test config."""
from fastdeploy.cache_manager.v1.cache_controller import CacheController
config = get_default_test_fd_config()
config.cache_config.enable_prefix_caching = enable_prefix_caching
config.cache_config.num_cpu_blocks = num_host_blocks
config.cache_config.cache_dtype = "bfloat16"
config.model_config.num_hidden_layers = num_layers
config.model_config.dtype = "bfloat16"
return CacheController(config, local_rank=0, device_id=0)
def create_mock_device_cache_kvs_map(
num_layers: int = 4,
local_rank: int = 0,
device_id: int = 0,
num_blocks: int = 100,
num_heads: int = 32,
block_size: int = 64,
head_dim: int = 128,
dtype: str = "bfloat16",
):
"""Helper to create mock device cache_kvs_map."""
import paddle
cache_kvs_map = {}
for layer_idx in range(num_layers):
key_name = f"key_caches_{layer_idx}_rank{local_rank}.device{device_id}"
val_name = f"value_caches_{layer_idx}_rank{local_rank}.device{device_id}"
key_tensor = paddle.zeros([num_blocks, num_heads, block_size, head_dim], dtype=dtype)
val_tensor = paddle.zeros([num_blocks, num_heads, block_size, head_dim], dtype=dtype)
cache_kvs_map[key_name] = key_tensor
cache_kvs_map[val_name] = val_tensor
return cache_kvs_map
def create_mock_host_cache_kvs_map(
num_layers: int = 4,
local_rank: int = 0,
device_id: int = 0,
base_ptr: int = 1000000,
):
"""Helper to create mock host cache_kvs_map (with int pointers)."""
cache_kvs_map = {}
for layer_idx in range(num_layers):
key_name = f"key_caches_{layer_idx}_rank{local_rank}.device{device_id}"
val_name = f"value_caches_{layer_idx}_rank{local_rank}.device{device_id}"
cache_kvs_map[key_name] = base_ptr + layer_idx * 10000
cache_kvs_map[val_name] = base_ptr + layer_idx * 10000 + 5000
return cache_kvs_map
def setup_transfer_env(controller, num_layers=4):
"""Helper to set up device and host cache for transfer tests."""
device_cache = create_mock_device_cache_kvs_map(num_layers=num_layers)
controller._transfer_manager.set_cache_kvs_map(device_cache)
host_cache = create_mock_host_cache_kvs_map(num_layers=num_layers)
controller._transfer_manager.set_host_cache_kvs_map(host_cache)
# ============================================================================
# Initialization Tests
# ============================================================================
class TestCacheControllerInit(unittest.TestCase):
"""Test CacheController initialization."""
def test_init_creates_executor(self):
"""Test that ThreadPoolExecutor is created on init."""
from concurrent.futures import ThreadPoolExecutor
controller = create_cache_controller()
self.assertIsNotNone(controller._executor)
self.assertIsInstance(controller._executor, ThreadPoolExecutor)
def test_init_creates_transfer_manager(self):
"""Test that TransferManager is created on init."""
controller = create_cache_controller()
self.assertIsNotNone(controller._transfer_manager)
def test_init_no_singleton_layer_counter(self):
"""Test that LayerDoneCounter is NOT created as singleton on init (per-transfer design)."""
controller = create_cache_controller(num_layers=4)
# In the new design, _layer_counter is None initially, set per transfer
self.assertIsNone(controller._layer_done_counter)
def test_init_empty_pending_evict_counters(self):
"""Test that pending evict counters list is empty on init."""
controller = create_cache_controller()
self.assertEqual(len(controller._pending_evict_counters), 0)
# ============================================================================
# load_host_to_device Tests
# ============================================================================
def make_done_counter(num_layers=4):
"""Create a pre-completed LayerDoneCounter for use in mocks."""
from fastdeploy.cache_manager.v1.cache_utils import LayerDoneCounter
counter = LayerDoneCounter(num_layers)
counter.mark_all_done()
return counter
class TestLoadHostToDevice(unittest.TestCase):
"""Test load_host_to_device returns LayerDoneCounter."""
def setUp(self):
self.controller = create_cache_controller(num_layers=4)
setup_transfer_env(self.controller, num_layers=4)
@patch("fastdeploy.cache_manager.v1.cache_controller.CacheController._submit_swap_task")
def test_returns_layer_done_counter(self, mock_submit):
"""Test that load_host_to_device returns LayerDoneCounter."""
from fastdeploy.cache_manager.v1.cache_utils import LayerDoneCounter
mock_submit.return_value = make_done_counter()
meta = CacheSwapMetadata(
src_block_ids=[10, 11, 12],
dst_block_ids=[0, 1, 2],
src_type="host",
dst_type="device",
)
counter = self.controller.load_host_to_device(meta)
self.assertIsNotNone(counter)
self.assertIsInstance(counter, LayerDoneCounter)
@patch("fastdeploy.cache_manager.v1.cache_controller.CacheController._submit_swap_task")
def test_single_metadata_completes_successfully(self, mock_submit):
"""Test that single metadata task completes with success."""
def fake_submit(meta, **kwargs):
meta.success = True
return make_done_counter()
mock_submit.side_effect = fake_submit
meta = CacheSwapMetadata(src_block_ids=[10], dst_block_ids=[0])
counter = self.controller.load_host_to_device(meta)
# Counter is already done (pre-completed)
self.assertTrue(counter.is_all_done())
self.assertTrue(meta.success)
@patch("fastdeploy.cache_manager.v1.cache_controller.CacheController._submit_swap_task")
def test_wait_for_layer(self, mock_submit):
"""Test wait_for_layer returns when layer is done."""
mock_submit.return_value = make_done_counter()
meta = CacheSwapMetadata(src_block_ids=[10], dst_block_ids=[0])
counter = self.controller.load_host_to_device(meta)
# Counter is pre-completed, wait_for_layer should return True immediately
result = counter.wait_for_layer(0, timeout=5.0)
self.assertTrue(result)
self.assertTrue(counter.is_layer_done(0))
@patch("fastdeploy.cache_manager.v1.cache_controller.CacheController._submit_swap_task")
def test_multiple_metadata_creates_separate_counters(self, mock_submit):
"""Test that multiple CacheSwapMetadatas create separate counters."""
mock_submit.side_effect = lambda *a, **kw: make_done_counter()
meta1 = CacheSwapMetadata(src_block_ids=[10], dst_block_ids=[0])
meta2 = CacheSwapMetadata(src_block_ids=[11], dst_block_ids=[1])
counter1 = self.controller.load_host_to_device(meta1)
counter2 = self.controller.load_host_to_device(meta2)
# Each should have its own counter
self.assertIsNot(counter1, counter2)
def test_empty_src_block_ids_sets_error(self):
"""Test that empty src block IDs set error."""
meta = CacheSwapMetadata(src_block_ids=[], dst_block_ids=[0])
self.controller.load_host_to_device(meta)
self.assertFalse(meta.success)
self.assertIsNotNone(meta.error_message)
def test_empty_dst_block_ids_sets_error(self):
"""Test that empty dst block IDs set error."""
meta = CacheSwapMetadata(src_block_ids=[10], dst_block_ids=[])
self.controller.load_host_to_device(meta)
self.assertFalse(meta.success)
self.assertIsNotNone(meta.error_message)
@patch("fastdeploy.cache_manager.v1.cache_controller.CacheController._submit_swap_task")
def test_returns_immediately_non_blocking(self, mock_submit):
"""Test that load_host_to_device returns without blocking."""
def slow_submit(*args, **kwargs):
time.sleep(0.5)
return make_done_counter()
mock_submit.side_effect = slow_submit
meta = CacheSwapMetadata(src_block_ids=[10], dst_block_ids=[0])
start = time.time()
self.controller.load_host_to_device(meta)
elapsed = time.time() - start
# load_host_to_device calls _submit_swap_task synchronously (submit to executor),
# so elapsed includes the mock's 0.5s sleep. Assert it completes within 1s.
self.assertLess(elapsed, 1.0)
# ============================================================================
# evict_device_to_host Tests
# ============================================================================
class TestEvictDeviceToHost(unittest.TestCase):
"""Test evict_device_to_host returns LayerDoneCounter."""
def setUp(self):
self.controller = create_cache_controller(num_layers=4)
setup_transfer_env(self.controller, num_layers=4)
@patch("fastdeploy.cache_manager.v1.cache_controller.CacheController._submit_swap_task")
def test_returns_layer_done_counter(self, mock_submit):
"""Test that evict_device_to_host returns LayerDoneCounter."""
from fastdeploy.cache_manager.v1.cache_utils import LayerDoneCounter
mock_submit.return_value = make_done_counter()
meta = CacheSwapMetadata(src_block_ids=[0, 1], dst_block_ids=[10, 11])
counter = self.controller.evict_device_to_host(meta)
self.assertIsNotNone(counter)
self.assertIsInstance(counter, LayerDoneCounter)
@patch("fastdeploy.cache_manager.v1.cache_controller.CacheController._submit_swap_task")
def test_single_metadata_completes(self, mock_submit):
"""Test that eviction completes successfully."""
def fake_submit(meta, **kwargs):
meta.success = True
return make_done_counter()
mock_submit.side_effect = fake_submit
meta = CacheSwapMetadata(src_block_ids=[0, 1], dst_block_ids=[10, 11])
counter = self.controller.evict_device_to_host(meta)
self.assertTrue(counter.is_all_done())
self.assertTrue(meta.success)
# ============================================================================
# submit_swap_tasks Tests
# ============================================================================
class TestSubmitSwapTasks(unittest.TestCase):
"""Test submit_swap_tasks method returns LayerDoneCounter."""
def setUp(self):
self.controller = create_cache_controller(num_layers=4)
setup_transfer_env(self.controller, num_layers=4)
@patch("fastdeploy.cache_manager.v1.cache_controller.CacheController._submit_swap_task")
def test_submit_swap_tasks_returns_layer_done_counter(self, mock_submit):
"""Test submit_swap_tasks returns LayerDoneCounter for swap_in."""
from fastdeploy.cache_manager.v1.cache_utils import LayerDoneCounter
mock_submit.return_value = make_done_counter()
evict_meta = CacheSwapMetadata(src_block_ids=[0], dst_block_ids=[10])
swap_in_meta = CacheSwapMetadata(src_block_ids=[10], dst_block_ids=[0])
counter = self.controller.submit_swap_tasks(evict_meta, swap_in_meta)
self.assertIsNotNone(counter)
self.assertIsInstance(counter, LayerDoneCounter)
@patch("fastdeploy.cache_manager.v1.cache_controller.CacheController._submit_swap_task")
def test_submit_swap_tasks_evict_only_returns_none(self, mock_submit):
"""Test submit_swap_tasks with only evict metadata returns None."""
mock_submit.return_value = make_done_counter()
evict_meta = CacheSwapMetadata(src_block_ids=[0], dst_block_ids=[10])
counter = self.controller.submit_swap_tasks(evict_meta, None)
# Evict-only returns None (no swap-in counter)
self.assertIsNone(counter)
@patch("fastdeploy.cache_manager.v1.cache_controller.CacheController._submit_swap_task")
def test_submit_swap_tasks_sets_swap_layer_done_counter(self, mock_submit):
"""Test submit_swap_tasks sets swap_layer_done_counter property."""
expected_counter = make_done_counter()
mock_submit.return_value = expected_counter
evict_meta = CacheSwapMetadata(src_block_ids=[0], dst_block_ids=[10])
swap_in_meta = CacheSwapMetadata(src_block_ids=[10], dst_block_ids=[0])
counter = self.controller.submit_swap_tasks(evict_meta, swap_in_meta)
# swap_layer_done_counter should be set
self.assertIs(self.controller.swap_layer_done_counter, counter)
# ============================================================================
# LayerDoneCounter Tests
# ============================================================================
class TestLayerDoneCounter(unittest.TestCase):
"""Test LayerDoneCounter independent sync primitive."""
def test_layer_done_counter_basic(self):
"""Test basic LayerDoneCounter functionality."""
from fastdeploy.cache_manager.v1.cache_utils import LayerDoneCounter
counter = LayerDoneCounter(num_layers=4)
# Initially not done
self.assertFalse(counter.is_all_done())
self.assertEqual(counter.get_completed_count(), 0)
# Mark one layer done
counter.mark_layer_done(0)
self.assertTrue(counter.is_layer_done(0))
self.assertFalse(counter.is_layer_done(1))
self.assertEqual(counter.get_completed_count(), 1)
self.assertFalse(counter.is_all_done())
def test_layer_done_counter_mark_all_done(self):
"""Test mark_all_done marks all layers."""
from fastdeploy.cache_manager.v1.cache_utils import LayerDoneCounter
counter = LayerDoneCounter(num_layers=4)
counter.mark_all_done()
self.assertTrue(counter.is_all_done())
self.assertEqual(counter.get_completed_count(), 4)
self.assertTrue(counter.is_layer_done(0))
self.assertTrue(counter.is_layer_done(3))
def test_layer_done_counter_wait_for_layer_immediate(self):
"""Test wait_for_layer returns immediately if done."""
from fastdeploy.cache_manager.v1.cache_utils import LayerDoneCounter
counter = LayerDoneCounter(num_layers=4)
counter.mark_all_done()
result = counter.wait_for_layer(0, timeout=1.0)
self.assertTrue(result)
def test_layer_done_counter_wait_all(self):
"""Test wait_all waits for all layers."""
from fastdeploy.cache_manager.v1.cache_utils import LayerDoneCounter
counter = LayerDoneCounter(num_layers=4)
# Mark all done
counter.mark_all_done()
result = counter.wait_all(timeout=1.0)
self.assertTrue(result)
self.assertTrue(counter.is_all_done())
def test_layer_done_counter_get_pending_layers(self):
"""Test get_pending_layers returns correct list."""
from fastdeploy.cache_manager.v1.cache_utils import LayerDoneCounter
counter = LayerDoneCounter(num_layers=4)
counter.mark_layer_done(1)
pending = counter.get_pending_layers()
self.assertEqual(pending, [0, 2, 3])
def test_layer_done_counter_callback(self):
"""Test callback is called on layer complete."""
from fastdeploy.cache_manager.v1.cache_utils import LayerDoneCounter
counter = LayerDoneCounter(num_layers=4)
callback_layers = []
def callback(layer_idx):
callback_layers.append(layer_idx)
counter.register_callback(callback)
counter.mark_layer_done(2)
self.assertEqual(callback_layers, [2])
def test_layer_done_counter_stats(self):
"""Test get_stats returns correct stats."""
from fastdeploy.cache_manager.v1.cache_utils import LayerDoneCounter
counter = LayerDoneCounter(num_layers=4)
counter.mark_layer_done(0)
counter.mark_layer_done(1)
stats = counter.get_stats()
self.assertEqual(stats["num_layers"], 4)
self.assertEqual(stats["completed_layers"], 2)
self.assertEqual(stats["pending_layers"], 2)
# ============================================================================
# Statistics Tests
# ============================================================================
class TestStats(unittest.TestCase):
"""Test statistics functionality."""
def test_get_stats_returns_expected_keys(self):
"""Test get_stats returns expected keys."""
controller = create_cache_controller(num_layers=4)
stats = controller.get_stats()
self.assertIn("initialized", stats)
self.assertIn("num_layers", stats)
self.assertTrue(stats["initialized"])
self.assertEqual(stats["num_layers"], 4)
# ============================================================================
# Reset Tests
# ============================================================================
class TestReset(unittest.TestCase):
"""Test reset_cache method."""
def setUp(self):
self.controller = create_cache_controller(num_layers=4)
setup_transfer_env(self.controller, num_layers=4)
@patch("fastdeploy.cache_manager.v1.cache_controller.CacheController._submit_swap_task")
def test_reset_cache_clears_pending_evict_counters(self, mock_submit):
"""Test reset_cache clears pending evict counters."""
mock_submit.return_value = make_done_counter()
evict_meta = CacheSwapMetadata(src_block_ids=[0], dst_block_ids=[10])
counter = self.controller.evict_device_to_host(evict_meta)
# Manually add counter to pending evict counters (simulating what submit_swap_tasks does)
self.controller._pending_evict_counters.append(counter)
self.assertEqual(len(self.controller._pending_evict_counters), 1)
result = self.controller.reset_cache()
self.assertTrue(result)
self.assertEqual(len(self.controller._pending_evict_counters), 0)
# ============================================================================
# KV Cache Management Tests
# ============================================================================
class TestKVCacheManagement(unittest.TestCase):
"""Test KV cache initialization and retrieval."""
def test_get_kv_caches_without_init(self):
"""Test get_kv_caches returns empty dict when not initialized."""
controller = create_cache_controller()
result = controller.get_kv_caches()
self.assertIsNotNone(result)
def test_get_host_cache_kvs_map_without_init(self):
"""Test get_host_cache_kvs_map returns empty dict when not initialized."""
controller = create_cache_controller()
result = controller.get_host_cache_kvs_map()
self.assertEqual(len(result), 0)
# ============================================================================
# Transfer Failure Tests
# ============================================================================
class TestTransferFailure(unittest.TestCase):
"""Test behavior when transfer fails."""
def setUp(self):
self.controller = create_cache_controller(num_layers=4)
setup_transfer_env(self.controller, num_layers=4)
@patch("fastdeploy.cache_manager.v1.cache_controller.CacheController._submit_swap_task")
def test_layer_by_layer_transfer_failure(self, mock_submit):
"""Test that transfer failure is properly reported via _submit_swap_task exception."""
def failing_submit(meta, **kwargs):
meta.success = False
meta.error_message = "CUDA error"
counter = make_done_counter()
return counter
mock_submit.side_effect = failing_submit
meta = CacheSwapMetadata(src_block_ids=[10], dst_block_ids=[0])
self.controller.load_host_to_device(meta)
# The error should be stored in meta.error_message
self.assertFalse(meta.success)
self.assertIsNotNone(meta.error_message)
self.assertIn("CUDA error", meta.error_message)
# ============================================================================
# Storage Placeholder Tests
# ============================================================================
class TestStoragePlaceholders(unittest.TestCase):
"""Test storage placeholder methods."""
def setUp(self):
self.controller = create_cache_controller(num_layers=4)
def test_prefetch_from_storage_returns_error_handler(self):
"""Test prefetch_from_storage returns error handler (not implemented)."""
from fastdeploy.cache_manager.v1.metadata import StorageMetadata
mock_metadata = MagicMock(spec=StorageMetadata)
handler = self.controller.prefetch_from_storage(mock_metadata)
self.assertIsNotNone(handler)
self.assertIsNotNone(handler.error)
def test_backup_device_to_storage_returns_error_handler(self):
"""Test backup_device_to_storage returns error handler (not implemented)."""
from fastdeploy.cache_manager.v1.metadata import StorageMetadata
mock_metadata = MagicMock(spec=StorageMetadata)
handler = self.controller.backup_device_to_storage([0, 1], mock_metadata)
self.assertIsNotNone(handler)
self.assertIsNotNone(handler.error)
def test_backup_host_to_storage_returns_error_handler(self):
"""Test backup_host_to_storage returns error handler (not implemented)."""
from fastdeploy.cache_manager.v1.metadata import StorageMetadata
mock_metadata = MagicMock(spec=StorageMetadata)
handler = self.controller.backup_host_to_storage([0, 1], mock_metadata)
self.assertIsNotNone(handler)
self.assertIsNotNone(handler.error)
class TestPDTransferPlaceholders(unittest.TestCase):
"""Test PD transfer placeholder methods."""
def setUp(self):
self.controller = create_cache_controller(num_layers=4)
def test_send_to_node_returns_error_handler(self):
"""Test send_to_node returns error handler (not implemented)."""
from fastdeploy.cache_manager.v1.metadata import PDTransferMetadata
mock_metadata = MagicMock(spec=PDTransferMetadata)
handler = self.controller.send_to_node(mock_metadata)
self.assertIsNotNone(handler)
self.assertIsNotNone(handler.error)
def test_wait_for_transfer_from_node_returns_error_handler(self):
"""Test wait_for_transfer_from_node returns error handler (not implemented)."""
from fastdeploy.cache_manager.v1.metadata import PDTransferMetadata
mock_metadata = MagicMock(spec=PDTransferMetadata)
handler = self.controller.wait_for_transfer_from_node(mock_metadata)
self.assertIsNotNone(handler)
self.assertIsNotNone(handler.error)
# ============================================================================
# CacheSwapMetadata Mapping Tests
# ============================================================================
class TestCacheSwapMetadataMapping(unittest.TestCase):
"""Test CacheSwapMetadata mapping property."""
def test_mapping_empty_when_not_success(self):
meta = CacheSwapMetadata(src_block_ids=[1, 2], dst_block_ids=[10, 11])
self.assertEqual(meta.mapping, {})
def test_mapping_returns_dict_after_success(self):
meta = CacheSwapMetadata(src_block_ids=[1, 2], dst_block_ids=[10, 11])
meta.success = True
expected = {1: 10, 2: 11}
self.assertEqual(meta.mapping, expected)
# ============================================================================
# write_policy Property Tests
# ============================================================================
class TestWritePolicy(unittest.TestCase):
"""Test write_policy property and related behavior."""
def test_write_policy_default(self):
"""Test write_policy reads from config."""
controller = create_cache_controller()
# Default config has write_policy set; just verify it's accessible
policy = controller.write_policy
self.assertIsInstance(policy, (str, type(None)))
def test_should_wait_for_swap_out_write_back(self):
"""Test _should_wait_for_swap_out returns True for write_back policy."""
from fastdeploy.cache_manager.v1.cache_controller import CacheController
config = get_default_test_fd_config()
config.cache_config.num_cpu_blocks = 50
config.model_config.num_hidden_layers = 4
config.cache_config.write_policy = "write_back"
controller = CacheController(config, local_rank=0, device_id=0)
self.assertTrue(controller._should_wait_for_swap_out())
def test_should_wait_for_swap_out_write_through(self):
"""Test _should_wait_for_swap_out returns False for write_through policy."""
from fastdeploy.cache_manager.v1.cache_controller import CacheController
config = get_default_test_fd_config()
config.cache_config.num_cpu_blocks = 50
config.model_config.num_hidden_layers = 4
config.cache_config.write_policy = "write_through"
controller = CacheController(config, local_rank=0, device_id=0)
self.assertFalse(controller._should_wait_for_swap_out())
# ============================================================================
# free_cache / free_gpu_cache Tests
# ============================================================================
class TestFreeCacheMethods(unittest.TestCase):
"""Test free_cache and free_gpu_cache methods."""
def setUp(self):
self.controller = create_cache_controller(num_layers=4)
setup_transfer_env(self.controller, num_layers=4)
def test_free_gpu_cache_clears_map(self):
"""Test free_gpu_cache clears the cache_kvs_map."""
device_cache = create_mock_device_cache_kvs_map(num_layers=4)
self.controller.cache_kvs_map = device_cache
self.assertGreater(len(self.controller.cache_kvs_map), 0)
self.controller.free_gpu_cache()
self.assertEqual(len(self.controller.cache_kvs_map), 0)
def test_free_cache_returns_true(self):
"""Test free_cache returns True on success."""
result = self.controller.free_cache()
self.assertTrue(result)
def test_free_gpu_cache_noop_when_empty(self):
"""Test free_gpu_cache is a no-op when cache_kvs_map is already empty."""
self.controller.cache_kvs_map = {}
# Should not raise
self.controller.free_gpu_cache()
self.assertEqual(len(self.controller.cache_kvs_map), 0)
if __name__ == "__main__":
unittest.main()