79 Commits

Author	SHA1	Message	Date
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
MingkunZhang	bb1f977c89	[Metax][Fix] add compilation option (#7209)	2026-04-07 02:43:43 -07:00
yzwu	8789329457	[Iluvatar] Support wi4a16 group_gemm (#7078)	2026-03-30 19:03:51 +08:00
Ding	defaffd5fb	【Hackathon 10th Spring No.45】FastDeploy 支持在 T4/V100 硬件的编译 -part (#6488) ... * fix(custom_ops): gate unsupported ops for sm70/sm75 build * fix(custom_ops): gate deepgemm exports to sm75+ only * [BugFix][OP] deduplicate CUDA sources to avoid moe_deepgemm multiple definition * revert two custom_ops files to 352f922f9	2026-03-23 19:16:23 +08:00
gongweibao	e4c9cac124	[BugFix] Cap nvcc -t threads to avoid compilation failures on high-co… (#6885) ... * [BugFix] Cap nvcc -t threads to avoid compilation failures on high-core machines On machines with many cores (e.g. 192), the nvcc -t flag was set to os.cpu_count(), causing each nvcc process to spawn that many internal threads. Combined with Paddle's ThreadPoolExecutor launching parallel compilations (also based on cpu_count), this leads to ~28K+ threads, resource exhaustion, and silent compilation failures. The linker then cannot find the missing .o files, but a second build succeeds because already-compiled objects are cached. Cap nvcc -t at 4 to keep total parallelism reasonable. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * Potential fix for pull request finding Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com> --------- Co-authored-by: gongweibao <gognweibao@baidu.com> Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com> Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>	2026-03-17 19:27:45 +08:00
yzwu	901b38c936	[Iluvatar] Optimize decode group_gemm and Support cuda graph for ernie (#6803)	2026-03-12 19:21:17 +08:00
wangyifei	b57c960837	cuda13.0, implement changes to CCCL (#6751)	2026-03-10 16:47:02 +08:00
AIbin	c3aceb6bdc	[Models][OP][Optimization] Support DeepSeek-v3.2 model, integrate DSA & Indexer architecture with FlashMLA/DeepGEMM (#6689) ... * Support DeepSeek-v3.2 model, integrate DSA & Indexer architecture with FlashMLA/DeepGEMM	2026-03-10 15:05:14 +08:00
sunxin	0dc7034ce0	[Model Runner] Deprecate not_need_stop (#6356) ... * Deprecate not_need_stop	2026-03-05 10:55:42 +08:00
yzwu	6674131b0b	[Iluvatar] Support CudaGraph and optimize flash_attn_unpadded and fused_neox_rope_embedding (#6553)	2026-03-02 14:07:17 +08:00
Yuanle Liu	6d3fede240	[OP][Feature] 统一 limit_thinking_content_length CUDA 算子，支持回复长度限制与注入序列 (#6493) ... * Initial plan * Migrate PRs #6311, #6129, #6305 to develop and merge unit tests Co-authored-by: yuanlehome <23653004+yuanlehome@users.noreply.github.com> * fix * update * fix * fix ci * fix ci * Initial plan * test: add test_chat_with_response_max_tokens to test_EB_VL_Lite_serving.py Co-authored-by: yuanlehome <23653004+yuanlehome@users.noreply.github.com> * test: add disable-thinking case to test_chat_with_response_max_tokens Co-authored-by: yuanlehome <23653004+yuanlehome@users.noreply.github.com> * test: add both reasoning_max_tokens and response_max_tokens case Co-authored-by: yuanlehome <23653004+yuanlehome@users.noreply.github.com> * fix ci * fix ci * fix ci --------- Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com> Co-authored-by: yuanlehome <23653004+yuanlehome@users.noreply.github.com>	2026-02-25 21:36:50 +08:00
yzwu	60e75ea8e8	[Iluvatar][CI] Fix cannot import get_stop (#6165)	2026-02-10 16:57:23 +08:00
chen	29a313a402	[Optimization] Support FA2/FA3/FA4 with attn_mask_q (#6354) ... * support FA4 sm100 * flash attn backend support mask * flash attn backend run flashmask correct * add test for flash_attn_backend and flash_attn_func * check * add test for fa4 * requirements.txt add fa4 whl * check test on sm100 * fix CI conflict * add enable_torch_proxy for flash_mask * lazy import fa4 * check * fix tests import * check test_load_mpt import	2026-02-05 14:39:00 +08:00
fxyfxy777	36547cfdb3	[Feature] FD_USE_PHI_FP8_QUANT (#6320) ... * add ut * add use_fd_quant env * rm mask_per_token_quant * add make ops list * USE_FD_FP8_QUANT -> FD_USE_PHI_FP8_QUANT 默认是true * modify comments * use bool type * Add function declaration	2026-02-03 22:33:03 -08:00
fxyfxy777	2ada119a38	[Optimize] optimize mask_quant & swiglu (#6222) ... * optimize mask_quant op speed up 1.5 * fix calculate sequence * add fused * rm log * push kernel code * add ut * accuracy ok * add ue8m0 * add ut * add merge develop * rm ut of mask_per_token_quant	2026-02-02 13:52:38 +08:00
xiaozude	030647521a	[Metax] adapt to the latest develop (#6282)	2026-01-29 23:21:20 -08:00
sunxin	27f8799f04	[Model Runner] Refactor execute_model for GPU async scheduling (#6176)	2026-01-28 14:19:33 +08:00
freeliuzc	49617d9832	[Feature]Support tag phase token enforce generation (#6034) ... * support tag phase token enforce generation * optimize note and some feature * fix sampler unit test --------- Co-authored-by: YuBaoku <49938469+EmmonsCurse@users.noreply.github.com>	2026-01-15 03:59:55 -08:00
MingkunZhang	7ad5737560	[Metax] adapt to gemm interface on different versions of maca (#5905) ... Co-authored-by: root <root@lt-wks-10-0-180-15.pub.metax-tech.com>	2026-01-07 10:02:24 +08:00
Neil Zhu	272a371635	[Metax] optimize flash attention backend (#5876)	2026-01-06 09:52:09 +08:00
MingkunZhang	f732d7d2ad	[Metax] adapt prefix caching & cpu swap (#5844) ... Co-authored-by: root <root@lt-wks-10-0-180-15.pub.metax-tech.com>	2025-12-31 17:02:48 +08:00
周周周	a3f0696e35	[BugFix] fix compile error in sm89 (#5809)	2025-12-29 16:55:52 +08:00
Juncai	412867fd99	[Feature] Support KV Cache Storage (#5571) ... * Support Mooncake Store * up * up * add op * fix conflict * fix error * up for comments * avoid thread lock * up * fix unittest * fix unittest * remove debug info * consider tp_size > 1 * add default rdma_nics * add utils * up * fix error --------- Co-authored-by: YuBaoku <49938469+EmmonsCurse@users.noreply.github.com>	2025-12-25 16:30:35 +08:00
yzwu	ac013803f3	[Iluvatar] Support V1_KVCACHE_SCHEDULER and paddleocr-vl rope mode (#5555)	2025-12-18 02:14:25 -08:00
FocusLuo	c3aaa7e441	[BugFix] Fixed build script issue on Intel HPU platforms (#5455) ... * [INTEL HPU] Fixed build script issue for non-gpu platforms Signed-off-by: Luo, Focus <focus.luo@intel.com> * [INTEL HPU] PR CI HPU will not use fixed version of fastdeploy_intel_hpu Signed-off-by: Luo, Focus <focus.luo@intel.com> --------- Signed-off-by: Luo, Focus <focus.luo@intel.com> Co-authored-by: YuBaoku <49938469+EmmonsCurse@users.noreply.github.com>	2025-12-11 16:36:37 +08:00
Neil Zhu	4403a21d4b	[Metax] refactor cutlass moe and optimize flash attention (#5361) ... * [Metax] refactor moe and flash attention backend --------- Co-authored-by: zhangchenyi_dl <16219492+zhangchenyidl@user.noreply.gitee.com>	2025-12-10 17:15:17 +08:00
xiaoxiaohehe001	6ca2651995	[Feature] Support noaux for eplb (#5143) ... * support noaux eplb * noaux_eplb * noaux_eplb * noaux_eplb	2025-11-21 14:10:32 +08:00
Neil Zhu	0edda75a56	[Metax] optimize cutlass moe and flash attention backend (#5128)	2025-11-20 16:12:35 +08:00
yzwu	3707af7a4f	[Iluvatar] add vl into ci and support v1 loader (#4774)	2025-11-11 10:50:17 +08:00
Neil Zhu	6de1ce3b25	[Metax] support ERNIE-4.5-VL-28B (#4820)	2025-11-07 04:55:49 -08:00
ming1753	cba185f1fe	[Feature] Optim PaddleOCR-VL (#4873) ... * [Feature] Optim PaddleOCR-VL * fix bug	2025-11-07 14:56:44 +08:00
Neil Zhu	c95d0740ec	[Metax] adapt cutlass moe for ernie-vl (#4685)	2025-11-03 17:44:27 +08:00
freeliuzc	11398790d3	[Speculative Decoding][MTP]Support attn mask offset (#4641) ... * [MTP]Merge support attn (#4591) * support mask_offset in speculate decoding * fix dummpy run output * add unit test * fix unit test import * support attn_mask_offset in mtp mode * add update_attn_mask op * fix unit test && fix code-style	2025-11-03 10:08:01 +08:00
ming1753	27746026c1	Skip building native architecture when specifying arch list (#4727)	2025-10-31 20:32:46 +08:00
Zhenghai Zhang	1712e1351b	【Hackathon 9th No.86】autogen MoeFastHardamardImplWrapper template_instantiation (#4592) ... * autogen MoeFastHardamardImplWrapper template_instantiation * fix codestyle * fix codestyle * add impl cu files	2025-10-30 10:28:36 +08:00
xiaozude	f7069b8057	[Metax] adapt DeepSeek (#4498)	2025-10-24 10:14:53 +08:00
RAM	775edcc09a	[Executor] Default use CUDAGraph (#3594) ... * add start intercept * Adjustment GraphOptConfig * pre-commit * default use cudagraph * set default value * default use cuda graph * pre-commit * fix test case bug * disable rl * fix moba attention * only support gpu * Temporarily disable PD Disaggregation * set max_num_seqs of test case as 1 * set max_num_seqs and temperature * fix max_num_batched_tokens bug * close cuda graph * success run wint2 * profile run with max_num_batched_tokens * 1.add c++ memchecker 2.success run wint2 * updatee a800 yaml * update docs * 1. delete check 2. fix plas attn test case * default use use_unique_memory_pool * add try-except for warmup * ban mtp, mm, rl * fix test case mock * fix ci bug * fix form_model_get_output_topp0 bug * fix ci bug * refine deepseek ci * refine code * Disable PD * fix sot yaml	2025-10-21 14:25:45 +08:00
Yuanle Liu	cef3164c3b	Optimizing the performance of think length limit using custom operators (#4279) ... * delete impl * delete min_length&max_length * support limit thinking content strategy * fix * fix * fix * update * fix set_value_by_flags_and_idx * fix * fix * fix * fix * update * fix * fix * fix typo * fix ci * fix * fix * support mtp * fix * fix * update * update	2025-10-20 21:09:13 +08:00
yzwu	4b661512ca	[Iluvatar GPU] Adapt VL model (#4313)	2025-10-17 16:13:38 +08:00
Zhenghai Zhang	6adfbe07ad	【Hackathon 9th No.86】autogen MultiQueryDecoderAttention template_instantiation -part (#4383) ... * split MultiQueryDecoderAttention template_instantiation * update comment * CI	2025-10-16 17:08:19 +08:00
Zhenghai Zhang	c46d5e48f8	【Hackathon 9th No.86】autogen MultiQueryAppendC8Attention template_instantiation -part (#4330) ... * split MultiQueryAppendC8Attention template_instantiation * update setup_ops.py * fix ci * fix bug	2025-10-10 15:07:48 +08:00
xiaozude	7c919070f7	[Metax] support cutlass moe & optimize flash attention (#4208)	2025-09-29 11:22:43 +08:00
李泳桦	6265f4385f	[feat] support prefix cache clearing when /clear_load_weight is called (#4008) ... * [feat] support clearing prefix cache (cherry-picked from release/2.1) * [fix] fix ipc suffix, use port instead * [fix] fix prefix caching not enabled * [fix] fix key/value_cache_scales indent * [fix] fix ep group all-reduce * [fix] fix clear/update lock not working when workers > 1 * [chore] add preemption triggered info log * [fix] fix code style * [fix] fix max_num_seqs config * [fix] do not force enable_prefix_caching=False in dynamic loading * [fix] fix ci * Revert "[fix] fix ci" This reverts commit 0bc6d55cc8. * [fix] initialize available_gpu_block_num with max_gpu_block_num * [fix] fix config splitwise_role * [fix] fix clearing caches synchronization and add more logs * [chore] print cache_ready_signal in log * [fix] fix scheduler_config.splitwise_role * [fix] fix cache_messager cache_ready_signal create=True * [fix] stop cache messager from launching in mixed deployment	2025-09-28 19:42:53 +08:00
fmiao2372	f1b5392e20	[Intel HPU] Support intel hpu platform (#4161) ... * [Intel HPU] Support intel hpu platform * fix some issues * apply precommit and move AttentionBackend_HPU * fix format issue * correct ops import * fix ci issue * update code in layers * fix code style issue * remove dense tp moe ep mode * fix enc_dec_block_num * fix rebase issue * rename hpu to gaudi in readme * rename ForwardMeta_HPU to HPUForwardMeta	2025-09-24 12:27:50 +08:00
yzwu	504461b6b5	[Iluvatar GPU] Optimize attention performance and fix moe load ckpt error (#3651)	2025-09-22 21:13:59 +08:00
YuanRisheng	d2d04c2d5e	[setup optimize]Support git submodule (#4033) ... * support git submodule * update setup * fix ci network * fix clone * revert clone linux * delete args * fix ci * update	2025-09-11 17:41:16 +08:00
zhupengyang	9d0074a91a	[xpu] add ep custom ops (#3911)	2025-09-10 12:22:50 +08:00
co63oc	5441538173	rename fused_get_rope.cu (#3752) ... * rename fused_get_rope.cu * fix * fix typos * fix * fix	2025-09-03 10:54:34 +08:00
co63oc	d6369b4d51	fix typos (#3684)	2025-09-01 17:50:17 +08:00
Sunny-bot1	fe5d09f9ee	[FIX]Fix Machete compile via ENABLE_MACHETE (#3727) ... * add ENABLE_MACHETE * fix * revert * update * pre_commit * fix * fix --------- Co-authored-by: Ayakouji <yuhongh@qq.com> Co-authored-by: Jiang-Jia-Jun <163579578+Jiang-Jia-Jun@users.noreply.github.com> Co-authored-by: aquagull <hongyuh@qq.com>	2025-08-30 17:50:17 +08:00