ViperEkura a5b238dd86 feat: add fp8 training via cublasLt dispatch ...

- fp8_mm kernel (csrc): cublasLt fp8 e4m3 gemm, TN layout mapped zero-copy
- custom::fp8_mm custom op: meta/cuda/cpu kernels + scale-corrected bf16 autograd
- aten::linear and linear_backward dispatch on CUDA key, zero model changes
- per-tensor scale or raw cast; single-GPU smoke loss matches bf16

2026-08-14 00:39:49 +08:00

..

config

feat: add persistent DataLoader workers

2026-08-09 11:38:50 +08:00

dataset

feat: add streaming IterableDataset for pretraining

2026-08-08 16:15:10 +08:00

extension

feat: add fp8 training via cublasLt dispatch

2026-08-14 00:39:49 +08:00

inference

feat: propagate attention backend across scheduler threads

2026-08-09 13:32:40 +08:00

model

refactor: split infer core into subpackages by concern

2026-08-08 23:43:05 +08:00

optim

fix: use torch.optim.AdamW in ManoAdamW instead of NAdamW

2026-08-01 09:20:54 +08:00

parallel

fix: broadcast ref/old model state_dict for FSDP

2026-07-31 08:32:22 +08:00

preprocessing

fix : correct prefill sampling and record alignment

2026-08-05 22:20:29 +08:00

serialization

refactor: remove H5 storage backend in favor of mmap bin

2026-07-29 12:50:27 +08:00

tokenize

refactor: simplify BaseFactory and separate ModelFactory from AutoModel

2026-07-30 09:38:20 +08:00

trainer

fix: report gradient snr in db

2026-08-09 13:40:27 +08:00

__init__.py

refactor: centralize logging in astrai.logging, replace ASTRAI_TIMED with log level

2026-08-08 12:39:27 +08:00

factory.py

refactor: simplify BaseFactory and separate ModelFactory from AutoModel

2026-07-30 09:38:20 +08:00

logging.py

refactor: split infer core into subpackages by concern

2026-08-08 23:43:05 +08:00

protocols.py

refactor: 重构训练后端为 Executor 模式

2026-05-24 20:35:44 +08:00

signal_handler.py

refactor: move signal_handler from parallel/ to top-level for broader reuse

2026-07-28 10:36:17 +08:00