microsoft/DeepSpeed

Pull Requests Commits

adding plotting script

mrwyattii committed 2 years ago

eecbc49c

correctness and skip tests we know will fail

mrwyattii committed 2 years ago

1dad6c43

expanding scripts to work with rest of tasks

mrwyattii committed 2 years ago

9214ed68

new benchmark script for multiple model families and tasks

mrwyattii committed 2 years ago

dbcb0d6b

Capture error message during sweep tests (#2351)

molly-smith committed 2 years ago

Verified eed40324

Refactor fused_bias_residual kernels for better readability (#2356)

arashb committed 2 years ago

Verified e14d40e5

Extend residual_add kernel tests to conver pre_attn_norm (#2354)

arashb committed 2 years ago

Verified 79692af1

Add missing pytest fixture scope (#2353)

arashb committed 2 years ago

Verified b450da4f

fix cuda invalid config error in dequant kernel (#2362)

GuanhuaWang committed 2 years ago

Verified 3486afb1

Update issue templates

jeffra committed 2 years ago

8e8c866d

Updated issue templates (#2363)

jeffra committed 2 years ago

Verified 70e883a1

Refactor gptj_residual_add kernels for better readability (#2358)

arashb committed 2 years ago

Verified 9df604bf

download cifar to blob storage (#2342)

mrwyattii committed 2 years ago

Verified 6ef16de1

docs(mixture-of-experts-inference): fix typo in tuto (#2345)

jqueguiner committed 2 years ago

Verified 2b1b0d2e

Add Onebit Optimzers in __init__ (#2340)

l4d2boomer committed 2 years ago

Verified f210256a

Kernel Data Conversion Utility (#2327)

cmikeh2 committed 2 years ago

Verified 9aa7b638

Inference profiling updates/fixes (#2348) (#2349)

awan-10 committed 2 years ago

Verified 99326438

fix zero docs (#2350)

jeffra committed 2 years ago

Verified 76de924b

Extend scratch buffer for long prompts (#2212)

cmikeh2 committed 2 years ago

Verified 3d097bb8

increase min pre-commit versions (#2346)

jeffra committed 2 years ago

Verified b76e0f4f

mem access for quantize kernel (#2331)

GuanhuaWang committed 2 years ago

Verified 954e0c61

Refactor residual add kernels (#2333)

arashb committed 2 years ago

Verified 48c5220b

MOE matmult with memaccess (#2336)

samadejacobs committed 2 years ago

Verified 12e1cb82

MOE residual matmult unit test (#2323)

samadejacobs committed 2 years ago

Verified 80b10d0c

jeffra committed 2 years ago

0f0a7a5d

Add more options to inference benchmark (#2325)

mrwyattii committed 2 years ago

Verified 15923810

only override forward if using cuda-graph (#2291)

jeffra committed 2 years ago

Verified cf638be9

add quant unit test (#2315)

GuanhuaWang committed 2 years ago

Verified 95d11517

refactor to use mem_access (#2317)

mrwyattii committed 2 years ago

Verified c199edac

ZeRO-Inference blog - Update README (#2322)

tjruwase committed 2 years ago

Verified 060078ab

Older