DeepSpeed
allreduce_always_fp16
#1487
Merged

allreduce_always_fp16 #1487

tjruwase merged 25 commits into deepspeedai:master from Dipet:allreduce_fp16
Dipet
Dipet fp16 allreduce
93a3c782
Dipet Dipet requested a review from awan-10 awan-10 4 years ago
Dipet Dipet requested a review from cli99 cli99 4 years ago
Dipet Dipet requested a review from conglongli conglongli 4 years ago
Dipet Dipet requested a review from eltonzheng eltonzheng 4 years ago
Dipet Dipet requested a review from jeffra jeffra 4 years ago
Dipet Dipet requested a review from minjiaz minjiaz 4 years ago
Dipet Dipet requested a review from niumanar niumanar 4 years ago
Dipet Dipet requested a review from RezaYazdaniAminabadi RezaYazdaniAminabadi 4 years ago
Dipet Dipet requested a review from samyam samyam 4 years ago
Dipet Dipet requested a review from ShadenSmith ShadenSmith 4 years ago
Dipet Dipet requested a review from tjruwase tjruwase 4 years ago
jeffra Merge branch 'master' into allreduce_fp16
5a1717fd
Dipet
jeffra Merge branch 'master' into allreduce_fp16
f815bf67
tjruwase
tjruwase commented on 2021-11-02
tjruwase
tjruwase commented on 2021-11-02
tjruwase
tjruwase commented on 2021-11-02
tjruwase Merge branch 'master' into allreduce_fp16
a38d147d
tjruwase
Dipet Undo sparse sum in nan check
09962464
tjruwase Merge branch 'master' into allreduce_fp16
87d953ad
Dipet Merge branch 'master' into allreduce_fp16
c7d24000
Dipet communication_data_type instead of fp32_allreduce and fp16_allreduce
32138682
Dipet sparse_allreduce with fp32 or fp16 data type
4ec370ca
tjruwase Merge branch 'master' into allreduce_fp16
d8c92fba
tjruwase
Dipet FIx communication_data_type checks
7a56b7a6
Dipet Merge branch 'allreduce_fp16' of github.com:Dipet/DeepSpeed into allr…
f92be8d9
Dipet
tjruwase Merge branch 'master' into allreduce_fp16
6ef30c5b
tjruwase
tjruwase commented on 2021-11-11
tjruwase
tjruwase commented on 2021-11-11
tjruwase Merge branch 'master' into allreduce_fp16
513bf003
tjruwase Merge branch 'master' into allreduce_fp16
41932438
tjruwase
tjruwase dismissed these changes on 2021-11-12
tjruwase tjruwase requested a review from tjruwase tjruwase 4 years ago
tjruwase tjruwase dismissed their stale review 4 years ago
Accidental
tjruwase Merge branch 'master' into allreduce_fp16
1289bfdf
Dipet Merge Master
2966952d
Dipet Allow only torch data types for communication_data_type
da8678e7
Dipet Merge branch 'master' into allreduce_fp16
658a380f
tjruwase
tjruwase commented on 2021-11-17
Dipet Fix Zero assert messages
d5200dc3
tjruwase Merge branch 'master' into allreduce_fp16
27db8cce
tjruwase
tjruwase approved these changes on 2021-11-17
tjruwase Merge branch 'master' into allreduce_fp16
34cfe969
tjruwase Merge branch 'master' into allreduce_fp16
423dbf30
tjruwase Merge branch 'master' into allreduce_fp16
6cf53a03
Dipet Merge branch 'master' into allreduce_fp16
00409037
tjruwase tjruwase merged d14baad9 into master 4 years ago

Login to write a write a comment.

Login via GitHub

Assignees
No one assigned
Labels
Milestone