| .. |
|
adapter_commons
|
2f61644f6e
SPMD optimizations (#824)
|
пре 3 месеци |
|
assets
|
411ac4f405
vlm: add support for Qwen2-VL model (#1015)
|
пре 2 месеци |
|
attention
|
4a7cb8f232
rocm: add custom paged attention kernels for ROCm (#1043)
|
пре 2 месеци |
|
common
|
055c8905a3
api: add sampling/engine option to return only deltas or final output (#1035)
|
пре 2 месеци |
|
compilation
|
0e5cf7f840
tpu: avoid dynamo guard eval overhead (#949)
|
пре 2 месеци |
|
distributed
|
61103b92d4
tpu: support single and multi-host TPUs on GKE and RayServe (#970)
|
пре 2 месеци |
|
endpoints
|
055c8905a3
api: add sampling/engine option to return only deltas or final output (#1035)
|
пре 2 месеци |
|
engine
|
05be6085ec
core: factor out input preprocessing into a separate class (#1039)
|
пре 2 месеци |
|
executor
|
f2b6dc3872
cpu: add support for W8A8 quantization via compressed-tensor (#1017)
|
пре 2 месеци |
|
inputs
|
05be6085ec
core: factor out input preprocessing into a separate class (#1039)
|
пре 2 месеци |
|
kv_quant
|
8a71788372
Add OLMoE (#772)
|
пре 4 месеци |
|
lora
|
bf4a4d8516
fix: do not register punica with torch if using older torch (#948)
|
пре 2 месеци |
|
modeling
|
6951928522
xpu: bump IPEX to 2.3, support GQA (#1042)
|
пре 2 месеци |
|
multimodal
|
411ac4f405
vlm: add support for Qwen2-VL model (#1015)
|
пре 2 месеци |
|
platforms
|
f2b6dc3872
cpu: add support for W8A8 quantization via compressed-tensor (#1017)
|
пре 2 месеци |
|
plugins
|
9797d38b24
torch.compile: allow adding custom compile backends via plugins (#1041)
|
пре 2 месеци |
|
processing
|
f561a54a43
core: fix async postprocessor in case of preemption (#1000)
|
пре 2 месеци |
|
prompt_adapter
|
30d02d0747
chore: remove peft as a requirement (#1006)
|
пре 2 месеци |
|
quantization
|
f2b6dc3872
cpu: add support for W8A8 quantization via compressed-tensor (#1017)
|
пре 2 месеци |
|
server
|
22a4cd4595
core: fix spec decode metrics and envs circular import (#889)
|
пре 2 месеци |
|
spec_decode
|
5c3b94de45
spec decode: move ops.advane_step to flash attention backend (#1005)
|
пре 2 месеци |
|
transformers_utils
|
411ac4f405
vlm: add support for Qwen2-VL model (#1015)
|
пре 2 месеци |
|
triton_utils
|
f1d0b77c92
[0.6.0] Release Candidate (#481)
|
пре 6 месеци |
|
worker
|
9797d38b24
torch.compile: allow adding custom compile backends via plugins (#1041)
|
пре 2 месеци |
|
__init__.py
|
f1d0b77c92
[0.6.0] Release Candidate (#481)
|
пре 6 месеци |
|
_core_ext.py
|
f1ea7711bd
core: do not compile ScalarType for torch < 2.4.0 (#938)
|
пре 2 месеци |
|
_custom_ops.py
|
4a7cb8f232
rocm: add custom paged attention kernels for ROCm (#1043)
|
пре 2 месеци |
|
_ipex_ops.py
|
6951928522
xpu: bump IPEX to 2.3, support GQA (#1042)
|
пре 2 месеци |
|
connections.py
|
c6c91edab7
ci: update & overhaul test units (#769)
|
пре 3 месеци |
|
constants.py
|
2f61644f6e
SPMD optimizations (#824)
|
пре 3 месеци |
|
py.typed
|
1c988a48b2
fix logging and add py.typed
|
пре 1 година |
|
scalar_type.py
|
f1d0b77c92
[0.6.0] Release Candidate (#481)
|
пре 6 месеци |
|
version.py
|
cbd51a208a
ci: bump to 0.6.5 (#964)
|
пре 2 месеци |
AlpinDale
4a7cb8f232
rocm: add custom paged attention kernels for ROCm (#1043)