Generation server using HF accelerate and DS inference by mayank31398 · Pull Request #321 · bigscience-workshop/Megatron-DeepSpeed

Commits on Jul 31, 2021

Commits on Aug 3, 2021

Create README.md
stas00
authored

Commits on Aug 4, 2021

Commits on Aug 5, 2021

Commits on Aug 6, 2021

Fix pretrain_gpt_single_node example script to have only one occurence of lr-decay-style
thomasw21
committed
better comment on TB writer (is_last_rank)
VictorSanh
committed

Commits on Aug 8, 2021

Add GLU variants (#47 )

jaketae
and
stas00
authored

Commits on Aug 10, 2021

[microsoft/Megatron-DeepSpeed sync] Commits including 2021-08-09 (#58 )

authored

Commits on Aug 11, 2021

use HuggingFace Datasets as source to build Megatron data files (#48 )

adammoody
and
thomasw21
authored

Commits on Aug 13, 2021

Commits on Aug 15, 2021

Commits on Aug 16, 2021

fix: doc_idx offset when merging indexed dataset files (#66 )
adammoody
authored

Commits on Aug 17, 2021

shuffle index list with numpy, scatter list, use file for large lists (#63 )

adammoody
and
thomasw21
authored

Commits on Aug 18, 2021

Commits on Aug 21, 2021

[testing] fixes for pt-1.10 (#71 )
stas00
authored

Commits on Aug 22, 2021

Commits on Aug 25, 2021

[codecarbon] integration (#15 )
stas00
authored

Commits on Aug 26, 2021

Commits on Aug 27, 2021

Commits on Sep 1, 2021

Commits on Sep 10, 2021

[requirements] fix format (#94 )
stas00
authored

Commits on Sep 13, 2021

[WIP] [codecarbon] sorting out CC warnings + logger preamble (#80 )
stas00
authored

Commits on Sep 15, 2021

Commits on Sep 16, 2021

Commits on Sep 20, 2021

Commits on Sep 21, 2021

Commits on Sep 24, 2021

ALiBi Implementation (#101 )

authored

Commits on Sep 29, 2021

[tests] flush std streams (#120 )

stas00
and
wade3han
authored

Commits on Sep 30, 2021

chore: update .gitignore
jaketae
authored

Commits on Oct 1, 2021

Commits on Oct 7, 2021

Save tokenizer in conversion script (#128 )

jaketae
and
stas00
authored

Commits on Oct 8, 2021

fix: only trigger ci on .py file changes (#131 )
jaketae
authored

Commits on Oct 10, 2021

Commits on Oct 11, 2021

[codecarbon] switch to master (#135 )
stas00
authored

Commits on Oct 18, 2021

Commits on Oct 19, 2021

Commits on Oct 20, 2021

backward compatibility for new chkpt keys (#147 )

stas00
and
conglongli
authored

Commits on Oct 21, 2021

fused softmax layer bug fix sync (#151 )

stas00
and
hyunwoongko
authored

Commits on Oct 22, 2021

Commits on Oct 25, 2021

Commits on Oct 26, 2021

Commits on Oct 27, 2021

[checkpoint] only one latest file (#164 )
stas00
authored

Commits on Oct 29, 2021

Commits on Nov 3, 2021

Commits on Nov 4, 2021

Commits on Nov 9, 2021

[CI] improvements (#185 )
stas00
authored

Commits on Nov 10, 2021

Commits on Nov 13, 2021

Full seqlen eval for CL+PP (#187 )
conglongli
authored

Commits on Nov 17, 2021

Support skip iteration flag (#177 )

jaketae
and
stas00
authored

Commits on Nov 18, 2021

add layernorm in Embedding (#191 )
stas00
authored

Commits on Nov 19, 2021

Commits on Nov 20, 2021

[CI] fix which tests get run (#199 )
stas00
authored

Commits on Nov 22, 2021

add missing space (#200 )
SaulLu
authored

Commits on Nov 23, 2021

Commits on Nov 24, 2021

[test] --partition-activations (#184 )
stas00
authored

Commits on Nov 25, 2021

chore: add tmp directory to .gitignore (#205 )
jaketae
authored

Commits on Nov 26, 2021

Commits on Nov 27, 2021

Revert "Checking we use fused kernels to compute scaled masked softmax on prefix lm (#209 )"
thomasw21
committed

Commits on Dec 1, 2021

Fix consumed_valid_samples counting for several valid dataloaders
TevenLeScao
authored

Commits on Dec 7, 2021

[TB] add throughput graphs (#210 )
stas00
authored

Commits on Dec 10, 2021

replay layer_norm_cuda_kernel.cu fixes (#216 )
stas00
authored

Commits on Dec 13, 2021

improve build (#207 )
stas00
authored

Commits on Dec 16, 2021

[logging] synced print (#217 )
stas00
authored

Commits on Jan 5, 2022

Commits on Jan 6, 2022

Commits on Jan 13, 2022

save args to txt file (#218 )
bhavitvyamalik
authored

Commits on Jan 15, 2022

Commits on Jan 21, 2022

Commits on Jan 24, 2022

fix add_scalar for pt<1.9 (#240 )
stas00
authored

Commits on Jan 28, 2022

Fix throughput unit (#241 )
janEbert
authored

Commits on Jan 29, 2022

[TB] log restarts (#234 )
stas00
authored

Commits on Feb 1, 2022

Alibi Tensor Parallel Fix (#244 )
DanielHesslow
authored

Commits on Feb 5, 2022

implement kill switch (#245 )
stas00
authored

Commits on Feb 8, 2022

--abort-on-unmet-fused-kernel-constraints (#247 )
stas00
authored

Commits on Feb 18, 2022

[apex FusedAdam] crash workaround (#249 )
stas00
authored

Commits on Feb 22, 2022

Replace approximate formula with exact one for throughput (#251 )
deepakn94
authored

Commits on Feb 25, 2022

Commits on Feb 26, 2022

Try to read the data path arguments directly from a file (#254 )
thomasw21
authored

Commits on Feb 28, 2022

Commits on Mar 1, 2022

Commits on Mar 7, 2022

Commits on Mar 9, 2022

switch to MixedFusedLayerNorm (#262 )
stas00
authored

Commits on Mar 11, 2022

Commits on Mar 18, 2022

Commits on Mar 25, 2022

torch.testing.assert_equal didn't make it (#273 )
stas00
authored

Commits on Apr 1, 2022

add stop alarm instructions
stas00
authored

Commits on Apr 13, 2022

Commits on Apr 25, 2022

Create CODEOWNERS
TevenLeScao
authored

Commits on Apr 26, 2022

Commits on May 3, 2022

Fix mixed fused layer norm to mimick nn.LayerNorm for torch>1.11 (#281 )
thomasw21
authored

Commits on May 31, 2022

[valid] deadlock workaround (#282 )
stas00
authored

Commits on Jun 5, 2022

Fix tflops glu computation (#283 )

Muennighoff
and
thomasw21
authored

Commits on Jun 22, 2022

Fix DS init (#285 )
Quentin-Anthony
authored

Commits on Jun 27, 2022

Commits on Jun 28, 2022

Eval harness (#212 )

authored

Commits on Jun 30, 2022

Merge MLM too fast 2 (#294 )
thomasw21
authored

Commits on Jul 2, 2022

MTF dataset and packing (#293 )

authored

Commits on Jul 4, 2022

CI fixes (#302 )
stas00
authored
sync layer norms (#272 )

stas00
and
thomasw21
authored

Commits on Jul 5, 2022

MTF train script (#295 )

authored

Commits on Jul 6, 2022

Add support for weighted train (#299 )
thomasw21
authored

Commits on Jul 7, 2022

Commits on Jul 10, 2022

Commits on Jul 11, 2022

fixes
stas00
committed
fix
stas00
committed

Commits on Jul 12, 2022

Commits on Jul 13, 2022

fix for offline use
stas00
committed
add benchmark
stas00
committed
add benchmark
stas00
committed
update
stas00
committed
cleanup
stas00
committed
update
stas00
committed
msecs
stas00
committed
cleanup
stas00
committed
improve
stas00
committed
fix benchmark, add warmup
stas00
committed
update
stas00
committed
fix; thanks Michael Wyatt
stas00
committed
clarify
stas00
committed
Merge branch 'bloom-inference' of https://github.com/bigscience-workshop/Megatron-DeepSpeed into bloom-inference
Reza Yazdani
committed
add bloom batch-inference script
Reza Yazdani
committed
removed the names :-)
Reza Yazdani
committed
fold the bs functionality from the other script
stas00
committed
fix
stas00
committed
restore do_sample
stas00
committed
dump generate args
stas00
committed

Commits on Jul 14, 2022

Commits on Jul 15, 2022

Commits on Jul 16, 2022

wip
stas00
committed
wip
stas00
committed

Commits on Jul 18, 2022

stats
stas00
committed

Commits on Jul 19, 2022

Commits on Jul 20, 2022

bug fix
stas00
committed
update
stas00
committed

Commits on Jul 22, 2022

update
stas00
committed

Commits on Aug 4, 2022

add server scripts

Mayank Mishra
authored and
Mayank Mishra
committed
fix bug

Mayank Mishra
authored and
Mayank Mishra
committed

Commits on Aug 6, 2022

new code

Mayank Mishra
authored and
Mayank Mishra
committed

Commits on Aug 7, 2022

working code

Mayank Mishra
authored and
Mayank Mishra
committed
fix bug

Mayank Mishra
authored and
Mayank Mishra
committed
update readme

Mayank Mishra
authored and
Mayank Mishra
committed
increase batch size for HF accelerate

Mayank Mishra
authored and
Mayank Mishra
committed
increase batch size

Mayank Mishra
authored and
Mayank Mishra
committed
support dynamic batch size with deepspeed

Mayank Mishra
authored and
Mayank Mishra
committed
drop num tokens

Mayank Mishra
authored and
Mayank Mishra
committed
drop return type

Mayank Mishra
authored and
Mayank Mishra
committed

Commits on Aug 8, 2022

oom

Mayank Mishra
authored and
Mayank Mishra
committed

Commits

Commits on Jul 28, 2021

Commits on Jul 29, 2021

Commits on Jul 30, 2021

Commits on Jul 31, 2021

Commits on Aug 3, 2021

Commits on Aug 4, 2021

Commits on Aug 5, 2021

Commits on Aug 6, 2021

Commits on Aug 8, 2021

Commits on Aug 10, 2021

Commits on Aug 11, 2021

Commits on Aug 13, 2021

Commits on Aug 15, 2021

Commits on Aug 16, 2021

Commits on Aug 17, 2021

Commits on Aug 18, 2021

Commits on Aug 21, 2021

Commits on Aug 22, 2021

Commits on Aug 25, 2021

Commits on Aug 26, 2021

Commits on Aug 27, 2021

Commits on Sep 1, 2021

Commits on Sep 10, 2021

Commits on Sep 13, 2021

Commits on Sep 15, 2021

Commits on Sep 16, 2021

Commits on Sep 20, 2021

Commits on Sep 21, 2021

Commits on Sep 24, 2021

Commits on Sep 29, 2021

Commits on Sep 30, 2021

Commits on Oct 1, 2021

Commits on Oct 7, 2021

Commits on Oct 8, 2021

Commits on Oct 10, 2021

Commits on Oct 11, 2021

Commits on Oct 18, 2021

Commits on Oct 19, 2021

Commits on Oct 20, 2021

Commits on Oct 21, 2021

Commits on Oct 22, 2021

Commits on Oct 25, 2021

Commits on Oct 26, 2021

Commits on Oct 27, 2021

Commits on Oct 29, 2021

Commits on Nov 3, 2021

Commits on Nov 4, 2021