microsoft/DeepSpeed

Pull Requests Commits

Workaround: use training kernels for BERT inference

eltonzheng committed 3 years ago

e56ea31d

Build DS inference wheel on Windows

eltonzheng committed 3 years ago

30381442

jeffra committed 3 years ago

220e5088

fix #2240: wrong time unit in flops_profiler (#2241)

yzs981130 committed 3 years ago

Verified 80f94c10

Update half precision header guards (#2261)

cmikeh2 committed 3 years ago

Verified 2a644488

Correctly detect CPU optimizer usage (#2257)

tjruwase committed 3 years ago

Verified cb5e05fe

Enable contiguous gradients with Z1+MoE (#2250)

siddharth9820 committed 3 years ago

Verified b288cf1b

jeffra committed 3 years ago

ebed51df

fix ds-inference without policy (#2247)

RezaYazdaniAminabadi committed 3 years ago

Verified c35bfe89

Make OPT policy backward compatible with pre-OPT transformers versions (#2254)

arashb committed 3 years ago

Verified fae896ef

Refactor dist tests: Checkpointing (#2202)

tjruwase committed 3 years ago

Verified 217338be

update videos (#2249)

jeffra committed 3 years ago

Verified 86164c48

Correctly detect zero_offload (#2213)

tjruwase committed 3 years ago

Verified a9b3bfa2

allow saving ckpt w/o ckpt json + bloom copy fix (#2237)

jeffra committed 3 years ago

Verified dce3acaa

Remove the random-generator from context during inference (#2228)

RezaYazdaniAminabadi committed 3 years ago

Verified fda63432

add doc for new bert example (#2224)

conglongli committed 3 years ago

Verified 5e42cc8b

Fix regression w. dist_init_required (#2225)

jeffra committed 3 years ago

Verified 7d8ad45d

fix typos in readme. (#2218)

zhjohnchan committed 3 years ago

Verified 9b418c1e

Add support of OPT models (#2205)

arashb committed 3 years ago

Verified 8b2a6371

[zero-3] print warning once and support torch parameter (#2127)

awan-10 committed 3 years ago

Verified b5ac0d54

use torch 1.9 (#2215)

jeffra committed 3 years ago

Verified f0054691

add cuda 11.7 (#2211)

jeffra committed 3 years ago

Verified a84f9da8

Correctly detect offload configuration (#2208)

tjruwase committed 3 years ago

Verified 5870f36c

fix table syntax (#2204)

kamalkraj committed 3 years ago

Verified 87b20133

Refactor Distributed Tests (#2180)

mrwyattii committed 3 years ago

Verified ac995198

Fix the tensor-slicing copy for qkv parameters (#2198)

RezaYazdaniAminabadi committed 3 years ago

Verified 8920308c

Log user config exactly (#2201)

tjruwase committed 3 years ago

Verified 28dfca8a

Adding additional instructiosn in the compression tutorial on pre-training distillation and quantization for GPT (#2197)

minjiaz committed 3 years ago

Verified f82846d7

Add gradient_average flag support for sparse grads (#2188)

Dipet committed 3 years ago

Verified b62e0cc5

Fix the layer-past for GPT based models (#2196)

RezaYazdaniAminabadi committed 3 years ago

Verified 54a9e1b9

Older