DeepSpeed
API for obtaining global gradient norm
#1292
Merged

API for obtaining global gradient norm #1292

tjruwase
FP16 fused and unfused grad norm query.
e35dd697
tjruwase Merge branch 'big-science' of github.com:microsoft/DeepSpeed into big…
f6b65ad0
tjruwase API for obtaining global unclipped gradient norm across parameter groups
5cda8e51
tjruwase tjruwase requested a review from ShadenSmith ShadenSmith 4 years ago
tjruwase tjruwase requested a review from awan-10 awan-10 4 years ago
tjruwase tjruwase requested a review from cli99 cli99 4 years ago
tjruwase tjruwase requested a review from conglongli conglongli 4 years ago
tjruwase tjruwase requested a review from eltonzheng eltonzheng 4 years ago
tjruwase tjruwase requested a review from jeffra jeffra 4 years ago
tjruwase tjruwase requested a review from minjiaz minjiaz 4 years ago
tjruwase tjruwase requested a review from niumanar niumanar 4 years ago
tjruwase tjruwase requested a review from RezaYazdaniAminabadi RezaYazdaniAminabadi 4 years ago
tjruwase tjruwase requested a review from samyam samyam 4 years ago
tjruwase tjruwase removed review request from samyam samyam 4 years ago
tjruwase tjruwase removed review request from conglongli conglongli 4 years ago
tjruwase tjruwase removed review request from awan-10 awan-10 4 years ago
tjruwase tjruwase removed review request from cli99 cli99 4 years ago
tjruwase tjruwase removed review request from eltonzheng eltonzheng 4 years ago
tjruwase tjruwase removed review request from minjiaz minjiaz 4 years ago
tjruwase tjruwase removed review request from RezaYazdaniAminabadi RezaYazdaniAminabadi 4 years ago
tjruwase tjruwase removed review request from niumanar niumanar 4 years ago
tjruwase tjruwase requested a review from samyam samyam 4 years ago
stas00
tjruwase Use global norm not group norms
dd02eee5
ShadenSmith
ShadenSmith approved these changes on 2021-08-09
tjruwase tjruwase merged cce85b89 into big-science 4 years ago
mrwyattii mrwyattii deleted the olruwase/global_gradient_norm branch 2 years ago

Login to write a write a comment.

Login via GitHub

Assignees
No one assigned
Labels
Milestone