ViperEkura
d08a92c7bd
feat: add frequency penalty to inference sampling pipeline
...
- Add FrequencyPenaltyStrategy (logit -= penalty * count)
- Per-task rep_window for penalty history lookup
- Wire through engine, task, executor, API layer
- Add --frequency_penalty and --rep_window to stream_chat.py
- 9 unit tests for frequency penalty strategy
2026-07-17 21:28:31 +08:00
ViperEkura
a1ea26d367
fix: rewrite GRPO data pipeline for offline record-level access
...
- process_list_field returns List[List[int]] preserving per-response boundaries
- GRPODataset rewritten to record-level __getitem__ (no windowing/stride)
- grpo_collate_fn pads variable-length responses into [B, G, R] tensors
- JsonlStore detects nested List[List[int]] and stores List[Tensor] per record
- Store._normalize skips nested-list keys from cumsum bookkeeping
- Pipeline._flush handles nested lists without cross-record flattening
- Export grpo_collate_fn from astrai.dataset
- 6 new GRPO tests + 2 updated builder tests, 114 total pass
2026-07-17 14:34:41 +08:00
ViperEkura
cd14d53707
feat: implement bfd_split packing strategy
...
- BFDSplitPacking splits over-length sequences into chunks before BFD
- All keys (loss_mask, position_ids, ...) split in lockstep for alignment
- No tokens lost vs bfd which truncates over-length sequences
- Tests: token preservation, chunk alignment, short unchanged, vs bfd
2026-07-17 12:38:56 +08:00
ViperEkura
e220413035
feat: support raw JSON files in dataset pipeline and JsonlStore
...
- detect_format now recognizes .json directories as jsonl store
- JsonlStore loads .json arrays and dicts alongside .jsonl
- tokenizer_path defaults to dataset dir when omitted
- Pipeline._iter_items handles .json files (arrays/single dict)
- Tests: detect_format, seq load, self-contained dataset dir
2026-07-17 12:20:03 +08:00
ViperEkura
84ed2327f5
feat: add --resume flag to decouple weight loading from training resumption
...
- Add --resume bool flag to train.py CLI
- --param_path always loads weights only by default
- --resume restores epoch, consumed_samples, optimizer & scheduler
- Checkpoint.load() now preserves full meta dict
- Update test_early_stopping to use new param_path/resume API
2026-07-16 14:23:23 +08:00
ViperEkura
2c7a71a9c0
refactor: separate old policy and ref model in GRPO strategy
...
- Split single ref_model into old_model (importance sampling ratio) and ref_model (frozen KL regularizer)
- Move ref_model/old_model creation from strategy __init__ to TrainContextBuilder, pass as explicit parameters
- Remove periodic sync_ref_model + sync_interval; add sync_old_model for external rollout loop to call
- DPOStrategy also receives ref_model from builder
- Fix std to use unbiased=False (population std per GRPO paper)
- Remove redundant tests (test_grpo_kl_zero_at_init, test_grpo_no_sync_interval_param)
- Remove --grpo_sync_interval CLI arg
2026-07-14 20:03:45 +08:00
ViperEkura
9bcd696580
fix: token-level ratio and prompt masking in GRPO strategy
...
- Mask prompt tokens to 0 so their logprobs excluded from ratio/KL
- Switch to token-level ratio + PPO clipping via reduction='none'
- Slice response token logprobs from full sequence output
- Replace k3 KL estimator with non-negative k1 estimator
- Fix epsilon from finifo.eps (~1e-38) to 1e-8
- Remove unused 'reduction' param from GRPOStrategy.__init__
- Clarify offline batch semantics in docstring
- Add 11 unit tests for masking, advantage, KL, sync, clipping
- Sync training.md and architecture.md docs
2026-07-12 21:24:09 +08:00
ViperEkura
841a582b28
refactor: split mask builder by single/multi output
...
- Extract SingleOutputMaskBuilder for SFT and pretrain configs
- Extract MultiOutputMaskBuilder for DPO and GRPO configs
- Keep SectionedMaskBuilder as backward-compatible facade
- Register "single" and "multi" names in MaskBuilderFactory
- Add parity and rejection tests for concrete builders
2026-07-08 21:18:34 +08:00
ViperEkura
5416c2e8fb
perf: replace paged KV cache with contiguous ContiguousCache, decode all groups
...
- Add KVCache/CacheView abstract base classes in cache.py
- Add ContiguousCache (contiguous per-slot buffer, default) alongside PageCache (paged, renamed from old KVCache)
- Merge make_table_tensor + bind into bind_tasks on KVCache interface
- Remove task_cached/task_record_hashes from base class (PageCache-only)
- Scheduler: decode all position groups instead of just the largest (eliminates 63% group skip rate)
- Scheduler: accept optional cache param for swapping implementations
- Model layer type hints use CacheView base class
- Batch 1-32: 1-7% speedup from eliminating Storage.gather overhead
- All 183 inference tests pass
2026-07-05 11:34:36 +08:00
ViperEkura
8999ca89b8
feat: add JSONL dataset store with on-the-fly tokenization
...
- Add JsonlStore registered under "jsonl" in astrai/dataset/storage.py
- Reuse PipelineConfig schema for JSONL dataset configuration
- Update detect_format to recognize JSONL directories and files
- Move save_h5/load_h5/save_bin/load_bin to astrai/serialization
- Split astrai/serialization.py into checkpoint/dataset submodules
- Add tests for JSONL detection, seq/SFT stores, and config roundtrip
2026-07-04 15:42:33 +08:00
ViperEkura
aabb0d83e9
refactor : replace iteration with consumed_samples
...
- Replace context.iteration with consumed_samples (global sample count)
- Add optimizer_step property derived from consumed_samples
- Checkpoint meta.json stores consumed_samples, drops iteration
- CLI --start_batch renamed to --start_samples (per-rank samples)
- Checkpoint dir naming: epoch_X_step_Y instead of epoch_X_iter_Y
- Metric log entries use step and consumed_samples fields
- Backward compat removed (old iteration checkpoints unsupported)
2026-06-30 18:42:42 +08:00
ViperEkura
25d4ea3f91
refactor : 压缩测试代码,消除重复
...
- fixture 替代重复实例化和 tokenizer 落盘
- parametrize 合并同构测试
- helper 消除 save_h5 + DatasetFactory.load 样板
- 净减 272 行
2026-06-19 14:54:39 +08:00
ViperEkura
6ae1828449
refactor : 清理工厂和配置系统中的死代码与冗余抽象
...
- 删除 Registry 中未使用的 category/priority 字段,_entries 简化为直接存储类引用
- 修正 __init_subclass__ 避免叶子类(AutoRegressiveLM 等)创建空注册表
- 删除 5 个工厂的薄 create() 覆写,统一使用 BaseFactory.create(name, *args, **kwargs)
- 删除 3 处零调用的 available_types/available_strategies 别名死代码
- 删除零调用的 BaseModelConfig.to_file 死代码
- 将 BaseConfig.from_json/to_json 重命名为 from_file/to_file,消除与子类重复
- 移除两个 inference builder 中总是被覆写的 prompt_tokens=0
2026-06-07 11:39:50 +08:00
ViperEkura
e7b18b7c03
refactor : BaseFactory 基类类型自动推导 + 移除冗余代码
...
- _validate_component 从 BaseFactory[T] 泛型参数自动解析基类类型,9 个子类覆写移除
- Registry 类内联到 BaseFactory._entries,移除未用的 list_by_category/list_by_priority
- _component_base 在 __init_subclass__ 时立即解析
- 数据集 4 个子类冗余 __init__ 移除
2026-06-06 21:23:41 +08:00
ViperEkura
9e31d4ef2b
feat : BaseToolParser.feed 增加可选 token_ids 参数
...
- format_chunk ABC 改为 (token, **kwargs),body/token_ids 通过 kw 传入
- ProtocolHandler._handle_stream 逐 token encode 并透传
- Anthropic builder 用 **kwargs 吸收不使用的参数,零变更
- 新增 3 个 token_ids 参数测试
2026-06-06 11:19:30 +08:00
ViperEkura
52aa4d01d5
feat : 推理层增加 vLLM 风格工具调用解析
...
- 新增 BaseToolParser 抽象基类,定义 feed/parse_complete 流式接口
- 新增 SimpleJsonToolParser,解析 {"name":"...","arguments":{...}} 格式
- 新增 ToolParserFactory,基于 BaseFactory 实现可插拔注册
- 集成 parser 到 OpenAIResponseBuilder,支持流式/非流式工具调用
- 扩展 ChatMessage 和 ChatCompletionRequest,增加 tools/tool_choice 字段
- 重构 format_chunk 接口,传入累积文本支持全量重新解析
- 新增 74 个单元测试,覆盖扫描/查找/流式解析/完整解析/工厂
2026-06-06 08:54:10 +08:00
ViperEkura
dc7d2cfbca
refactor : FastAPI 懒加载单例,消除模块级副作用
...
- import astrai.inference 不再在模块加载时创建 FastAPI 实例
- 路由移至 APIRouter;get_app() 首次调用时懒构造单例
- _create_engine 和 run_server 的 param_path 改为必填
- 更新测试改用 get_app() 替代模块级 app
2026-06-04 15:52:27 +08:00
ViperEkura
b36a78c612
test : SFT 测试数据补全 position_ids 字段
...
- dummy_data 添加 position_ids 匹配 required_keys
2026-06-04 14:01:04 +08:00
ViperEkura
02a7cb9fa0
feat : preprocessing 支持 DPO/GRPO 多输出格式
...
- InputConfig 新增 sources 字段驱动多输出映射
- SectionedMaskBuilder 提取 _process_sections/_build_multi 模板方法
- Pipeline 泛化 accumulate 逻辑处理多 key 结果
- 测试拆分为 config/builder/pipeline 三文件,纯函数风格
2026-06-03 10:32:10 +08:00
ViperEkura
01ce1fb9e3
refactor : Pipeline 去除去重,ids 重命名为 sequence,泛型透传
...
- 移除 Pipeline 内置去重逻辑及 dedup_signature 工具函数
- 删除 ProcessingConfig.deduplicate 字段
- builder 返回 'sequence' 替代 'ids',与 dataset 层统一
- pipeline 纯透传,泛型处理任意 key 补齐默认值
2026-05-31 15:14:27 +08:00
ViperEkura
dbe5891201
refactor : 统一 SectionedMaskBuilder,支持可配置 dtype
...
- 三合一 MaskBuilder,移除 chat/instruction/text,统一为 sections 配置
- OutputConfig 增加 dtype 字段 (per-key,默认 int32)
- 移除 from __future__ import annotations
- 测试适配新配置格式
2026-05-31 14:24:10 +08:00
ViperEkura
2a65c3314c
fix : 修复 created 时间戳、bin 多 shard 覆盖与文档遗漏
...
- openai.py/anthropic.py: created 从 0 改为 int(time.time())
- openai.py: ChatCompletionRequest 不支持参数非默认值时 warning
- pipeline.py: bin 多 shard 使用子目录避免静默覆盖
- storage.py: MmapStore/detect_format 支持多 shard 聚合加载
- architecture.md: mermaid 类图新增 Pipeline 类
- preprocessing.md: 新增多 shard 输出布局与 Python API 示例
- protocol.py: docstring "6 methods" 改为 "5 methods"
2026-05-30 23:03:42 +08:00
ViperEkura
31ae2deeba
refactor : BaseConfig 提供 from_json/to_json,嵌套 config 自动反序列化
...
- from_json/to_json 上提至 BaseConfig,所有子类自动继承
- _coerce 新增 dict 到 BaseConfig 子类的递归反序列化,消除子类 from_dict 重载
- PipelineConfig 等子类仅声明字段,零样板代码
- 测试 tokenizer 改为自包含 BPE(含 chat template),不依赖 params/ 目录
- 特殊 token 改用 ASCII 字符,兼容所有平台
2026-05-30 21:04:19 +08:00
ViperEkura
69207e2c57
refactor : 基于声明式 JSON 配置的预处理管线重构
...
- 用工厂注册的 MaskBuilder(chat/instruction/text)替换硬编码的 _transform_* 方法
- mask 规则以 role-to-action 映射声明在配置中,与 chat_template 完全解耦
- 单次编码 + role-span 追踪替代两次编码 + 长度差计算 mask 的方式
- 支持多轮对话训练:所有 assistant 轮次参与训练,而非仅最后一轮
- 新建 astrai.preprocessing 包(builder.py + pipeline.py),删除 astrai/preprocess.py
- CLI 精简为 --config 参数,所有参数通过 PipelineConfig JSON 配置
- 新增 PipelineConfig、InputConfig、ProcessingConfig、OutputConfig dataclass
- 文档:assets/docs/preprocessing.md
- 27 个测试覆盖 mask builder、pipeline、配置序列化、工厂注册
2026-05-30 20:45:09 +08:00
ViperEkura
d4451f6afb
fix : 并行训练 state_dict 收集与训练/推理并发缺陷
...
- FSDPExecutor: unwrap_model 返回全量 state_dict (state_dict_type FULL);use_orig_params=True
- DDPExecutor/BaseExecutor: unwrap_model 统一返回 model.module.state_dict() / model.state_dict()
- CheckpointCallback: 走 executor.unwrap_model 拿完整 state_dict
- strategy.py: 移除 FSDP/DDp 依赖;create_ref_model(model_fn, state_dict) 纯函数
- TrainContextBuilder: 传递 model_fn + executor 到 strategy
- GRPOStrategy.sync_ref_model: 通过 executor.unwrap_model 获取完整权重
- TaskManager.wait_for_tasks: 锁内检查队列,消除 clear/set 竞态
- ProtocolHandler: stop token 不再计入 completion_tokens(流式/非流式)
2026-05-29 21:12:52 +08:00
ViperEkura
7c99da155c
refactor: 删除数据流中的 JSONStore
...
- 移除 JSONStore 及相关函数,训练框架不再依赖 tokenizer
- Store 层只保留 H5Store 和 MmapStore 两种后端
2026-05-28 15:54:26 +08:00
ViperEkura
629e72385b
fix : 修复存储层 bug,JSON 切换为 JSONL,补齐测试覆盖
...
- save_bin/load_bin: save_json/load_json 替换为直接 json.dump/json.load,修复致命 bug
- _normalize: 空 cum 列表 guard,防止 IndexError
- load_json: 改为仅支持 JSONL 逐行解析 (json.loads),移除 .json 支持
- detect_format: 只匹配 *.jsonl,不再匹配 *.json
- save_json: 输出扩展名改为 .jsonl
- GRPODataset.__getitem__: 补齐 .to(dtype=torch.long/bool) 与其他数据集一致
- load_bin: np.memmap mode='r+' 消除 PyTorch 不可写 tensor 警告
- 新增 16 个测试: bin roundtrip, mmap load, 空 key, JSONL 多行/文本, GRPO dtype/load, detect_format bin/jsonl, fetch multi-key/越界, json_to_bin 转换, DPO from JSONL, 显式 storage_type
2026-05-28 15:29:46 +08:00
ViperEkura
6e150ea6d0
refactor : Storage 层重构为 Store,移除 Fetcher 中间层,支持多段数据与显式长度
...
- 合并 BaseStorage + MultiSegmentFetcher + BaseSegmentFetcher 三层为 Store ABC
- Store._data 直接持有 Dict[str, List[Tensor]],不做强制拼接避免 OOM
- _fetch_key 统一用 bisect 跨段切片,单段多段同一路径
- _length 显式存储(min total across keys),__len__ 返回 O(1)
- MmapStore/H5Store/JSONStore 统一走 _normalize() 注册分段并预计算累积长度
- 所有 I/O 函数 (save_h5/load_h5/json_to_bin 等) 保持不变
2026-05-28 14:23:49 +08:00
ViperEkura
4145d35e3c
refactor: 检查点加载重构,路径替代对象传递
...
- model: nn.Module -> model_fn 工厂函数,spawn 边界只传字符串
- Trainer.train(resume_dir=path) — Checkpoint 不再通过 pickle 传递
- TrainContextBuilder.with_resume_dir(path) — 自动检测 meta.json 分流 resume/from-scratch
- CheckpointCallback: 拆分 state_dict 收集(全 rank)与磁盘写入(rank-0),修复 FSDP 死锁
- serialization: load_torch 支持 broadcast,消除 _load_extra/_load_torch_broadcast
- optimizer/scheduler 恢复逻辑内联到 build(),在 executor.prepare() 之后执行
- pyproject.toml: ruff exclude build/ 避免 CI 扫描构建产物
2026-05-27 20:15:29 +08:00
ViperEkura
65ab69543b
refactor: 统一序列化层,消除分散的 I/O 路径
...
- Checkpoint 改为 @dataclass,内聚 save/load 方法
- 提取 save_safetensors/load_safetensors/save_json/load_json 共享工具
- 新增 save_model/load_model_config/load_model_weights 模块函数
- automodel 和 lora 统一委托到 serialization 模块
2026-05-26 16:44:40 +08:00
ViperEkura
dd1b39f435
fix: ProgressBar默认输出到stdout
...
- file参数默认值改为None, 内部用 or sys.stdout 兜底
- 清理inference API中未使用的import (Optional, time, field)
- 删除test_protocol中未使用的ctx变量
2026-05-26 13:27:05 +08:00
ViperEkura
94d6e713e9
test: 补充推理协议层单测覆盖
...
- StopChecker、GenContext、StopInfo 单测
- OpenAIResponseBuilder / AnthropicResponseBuilder 全部方法
- Anthropic 停止序列裁剪逻辑(含 unyielded 边界)
- GenerationRequest 参数校验含负值边界
- Scheduler prefill 短路验证
2026-05-26 00:21:52 +08:00
ViperEkura
a4688021bf
feat: 新增LoRA微调模块
...
- LoRALinear基于register_parameter托管base weight,state_dict路径不变
- inject_lora/merge_lora/save_lora/load_lora完备封装
- 24个单元测试覆盖注入、合并、存取、边界场景
2026-05-25 20:15:31 +08:00
ViperEkura
7fa69572c0
fix: 测试日志写入临时目录避免冗余文件
2026-05-24 20:54:59 +08:00
ViperEkura
785d65436c
fix: 修复 to_dict list 类型丢失与 OpenAI stop 参数失效
...
- to_dict() 增加 list 类型序列化支持,metrics 等字段不再丢失
- OpenAIHandler 补充 get_stop_sequences/on_token,读取 request.stop 并检测停止序列
- 文档类图补充缺失字段、修正关系分类、ChatCompletionRequest 字段增加 Optional
2026-05-19 21:07:07 +08:00
ViperEkura
2c2697390d
feat: 新增 GradientCheckpointingCallback
...
- TrainConfig.gradient_checkpointing_modules 指定模块类型
- apply 递归遍历,兼容 DDP,不硬编码模型结构
- modules=None 时静默跳过,零开销
2026-05-17 18:21:05 +08:00
ViperEkura
97c7ac0f4f
refactor: Transformer更名为AutoRegressiveLM并新增EmbeddingEncoder
...
- AutoRegressiveLM 注册名改为 autoregressive_lm
- 新增 EmbeddingEncoder 支持 mean/cls/last pooling
- ModelConfig 增加 pooling_type / normalize_embeddings 字段
- 导入、注释、测试全部同步更新
2026-05-17 15:29:20 +08:00
ViperEkura
a44fd22a99
fix: 修复训练与模型参数传递问题
...
- state_dict_fn 传入 CheckpointCallback,修复多卡 DDP 下 key 前缀丢失
- MLA 增加 use_qk_norm 支持,消除参数静默丢失
- moe_topk_method 统一命名为 topk_method
- checkpoint 回调移至最前
2026-05-17 11:20:13 +08:00
ViperEkura
f91bfee33e
refactor: Config序列化统一BaseConfig基类
...
- 新增astrai/config/base.py,提供to_dict/from_dict基类
- 统一命名:load/save → from_file/to_file
- Checkpoint.meta合并训练配置到meta.json
- sys.stderr.warn → warnings.warn
- from_file改为classmethod
2026-05-16 22:06:39 +08:00
ViperEkura
d7a7f570ed
refactor: 训练循环改为两重迭代并统一参数命名
...
- 训练循环从三重(epoch→batched→batch)改为二重(epoch→batch)
- batch_size → batch_per_device, accumulation_steps → grad_accum_steps
- scheduler 移入 step block 对齐 optimizer 更新步
- GradientClippingCallback 改用 on_step_begin 避免零梯度裁剪
- 移除 _train_impl 误导性的 -> Checkpoint 标注
- total_steps 修除为向下取整并精简为一行
- warmup_steps 改为 warmup_ratio (默认0.05)
2026-05-16 21:27:35 +08:00
ViperEkura
7dea929788
refactor: checkpoint 按 HF 方式存独立 .pt 文件,callback 接管恢复
...
- Checkpoint.save/load: extra 逐 key 写为 {key}.pt 而非单个 extra.pt
- meta.json 新增 timestamp
- CheckpointCallback: save_extra/load_extra 静态方法 + extra_keys 类属性
- on_train_begin 接管 optimizer/scheduler 恢复,TrainContextBuilder 不再传 load_extra_fn
2026-05-16 18:29:04 +08:00
ViperEkura
04c0dc7a47
refactor: Storage 改用工厂模式,server reload 接入 uvicorn
...
- 新增 StorageFactory(BaseFactory[BaseStorage]) 替代手写 dict 注册
- H5Storage / JSONStorage 通过 @StorageFactory.register 注册
- dataset.py 使用 StorageFactory.create() 替代 create_storage()
- 删除 create_storage / available_storage_types 死函数
- server.py reload 参数正式传入 uvicorn.run()
2026-05-16 17:00:26 +08:00
ViperEkura
0ba8c70ce1
fix: 修复 MLA 多个 bug 并缩小测试模型参数
...
- MLA kv_b_proj 输出维度和 q_rope 切分偏移修复
- 打通 MLA 配置从 ModelConfig 到 DecoderBlock 的传递路径
- rope_theta 配置不再被忽略,MLA 使用 qk_rope_head_dim
- tie_weight 使用 is True 避免 None 隐式生效
- norm_eps/rope base 类型标注修正
- 测试模型参数缩小 (dim=8, head_dim=4)
- 新增 6 种架构配置 × 2 场景的前向传播测试
2026-05-16 14:57:43 +08:00
ViperEkura
e3382f6bb5
fix: 修复推理引擎 batch decode 中多项正确性与并发问题
...
- scheduler: decode 分组由幂次分桶改为精确 next_pos,消除 KV cache 位置错乱
- task: activate() 加锁操作 active_tasks,消除数据竞争
- engine: wait_completion 加超时,防止分配失败时永久死锁
- sample: TopKStrategy 向量化为 per-sample threshold,尊重各 task 的 top_k
- cache: Storage.write/gather 中 -1 页改用 mask 处理,防数据污染
- executor: prefill 逐 task 循环改为单次 tensor 调用
2026-05-14 21:31:39 +08:00
ViperEkura
205b40bd28
refactor: 重构 cache 和 inference 参数体系,分离存储与分配
...
- 合并 GenerationRequest/GenerationParams,统一 max_tokens 参数名
- PagePool/PrefixCache 分离为 Allocator + PrefixCache + PagePool
- 拆分 KV 存储为独立 Storage 类,PagedCache → KVCache,CacheView → KvcacheView
- Allocator.inc_ref 移除 LRU 防止竞争,Storage.write 增加负页防御
- Allocator/PrefixCache/TaskTable 加 threading.Lock 保证线程安全
- server.py uvicorn.run 改为传 app 对象修复导入错误
- benchmark.py 适配 KVCache 新 API
2026-05-14 20:05:08 +08:00
ViperEkura
2196c34c52
refactor: 重构 inference 模块架构,引入设计模式并分组文件
...
- 新增 protocol.py 协议层,Template Method 模式消除流/非流分支 45% 重复
- SSEBuilder 统一 SSE 构造,StopChecker 独立 stop_sequence 检测
- AnthropicHandler 追踪已产出文本,修复 stop 时重复 delta
- server.py 路由从约 100 行缩减至 3 行
- 拆分为 core/(cache/executor/scheduler/task)和 api/(protocol/server)
- 外部保持二级导入路径(from astrai.inference import Name)
- 删除所有分隔线注释,代码按语义自然分组
2026-05-14 17:42:37 +08:00
ViperEkura
6d6ef99e66
perf: 消除 PagedCache.write 中的 position_ids GPU 同步,解码提速 15%
...
- CacheView.write 用 total_len - k.size(1) 推导 start_pos,替代 position_ids[0,0].item()
- 移除 GQA/MLA/DecoderBlock 中不再使用的 position_ids 参数
- PagedCache.write 参数 position_ids:Tensor → start_pos:int
2026-05-14 15:37:48 +08:00
ViperEkura
c0effc9f5b
refactor: 位置编码改用 position_ids [B,S],简化 attention mask 构建
...
- RotaryEmbedding/CacheView 接受 position_ids 替代 start_pos
- process_attention_mask 用 position_ids >= arange 做逐位置 causal
- 训练/无 KV cache 时 position_ids=None 内部自动处理
- 移除 executor/benchmark 中冗余的 input_mask 构造
2026-05-14 13:26:31 +08:00
ViperEkura
df0845e916
chore: 解耦 Executor/Scheduler/TaskManager,修复 stop 页泄漏,移除 ServerState 全局单例
2026-05-12 13:47:55 +08:00
ViperEkura
7440e9c809
style: 重命名 test_scheduler_concurrency 为 test_scheduler
2026-05-12 12:24:36 +08:00