Skip to main content
darkbird
Projects
psy-llama-pretraining
Log in
Sign up
Project
Workspace
Runs
Automat.
Sweeps
Reports
Artifacts
Kirp's workspace
Personal workspace
Automated workspace
Changes are only visible to you.
Runs
4
Name
3 visualized
Train with default qkvogdu+lm+emb on 4*A6000 lr=1e-4 pretraining
Train with default qkvogdu+lm+emb on 4*A6000 lr=1e-4 pretraining
Train with t10000 map qkvogdu+lm+emb on 8*A6000 lr=1e-4 pretraining
Train with t10000 map qkvogdu+lm+emb on 8*A6000 lr=1e-4 pretraining
Train with t10000 random qkvogdu+lm+emb on 4*A6000 lr=1e-4 pretraining
Train with t10000 random qkvogdu+lm+emb on 4*A6000 lr=1e-4 pretraining
Train with default qkvogdu
Train with default qkvogdu
1-4
of 4
Previous
Next
eval/steps_per_second
eval/steps_per_second
50
100
150
200
250
300
350
Step
0.08
0.1
0.12
0.14
0.16
0.18
0.2
Train with default qkvogdu+lm+emb on 4*A6000 lr=1e-4 pretraining
Train with t10000 map qkvogdu+lm+emb on 8*A6000 lr=1e-4 pretraining
Train with t10000 random qkvogdu+lm+emb on 4*A6000 lr=1e-4 pretraining