| .. |
|
__init__.py
|
07aa2a492f
upstream: add option to specify tokenizer
|
1 year ago |
|
block.py
|
7df7b8ca53
optimization: reduce end-to-end overhead from python obj allocation (#666)
|
6 months ago |
|
config.py
|
5bec8fbb1b
tpu: add support for async postprocessing (#968)
|
2 months ago |
|
connections.py
|
f1d0b77c92
[0.6.0] Release Candidate (#481)
|
6 months ago |
|
envs.py
|
0e5cf7f840
tpu: avoid dynamo guard eval overhead (#949)
|
2 months ago |
|
grammar.py
|
8a71788372
Add OLMoE (#772)
|
5 months ago |
|
logger.py
|
22a4cd4595
core: fix spec decode metrics and envs circular import (#889)
|
2 months ago |
|
logits_processor.py
|
8a71788372
Add OLMoE (#772)
|
5 months ago |
|
outputs.py
|
0e558e9b2f
fix: loading chameleon model with TP>1 (#695)
|
6 months ago |
|
pooling_params.py
|
2f61644f6e
SPMD optimizations (#824)
|
3 months ago |
|
sampling_params.py
|
8d9f1fd4e6
feat: add single user mode (#927)
|
2 months ago |
|
sequence.py
|
0dfa6b60ec
core: support logprobs with multi-step scheduling (#963)
|
2 months ago |
|
test_utils.py
|
8a71788372
Add OLMoE (#772)
|
5 months ago |
|
utils.py
|
f1ea7711bd
core: do not compile ScalarType for torch < 2.4.0 (#938)
|
2 months ago |
AlpinDale
5bec8fbb1b
tpu: add support for async postprocessing (#968)