Hugging Face
Models
Datasets
Spaces
Posts
Docs
Solutions
Pricing
Log In
Sign Up
DUAL-GPO
/
phi-2-gpo-iter-1
like
0
PEFT
Safetensors
HuggingFaceH4/ultrafeedback_binarized
phi
alignment-handbook
Generated from Trainer
trl
dpo
custom_code
License:
mit
Model card
Files
Files and versions
Community
Use this model
dd5eeed
phi-2-gpo-iter-1
/
train_results.json
lole25
Model save
44ca9b5
verified
4 months ago
raw
history
blame
No virus
197 Bytes
{
"epoch"
:
1.98
,
"train_loss"
:
0.010306676939850854
,
"train_runtime"
:
2255.3236
,
"train_samples"
:
61135
,
"train_samples_per_second"
:
0.887
,
"train_steps_per_second"
:
0.055
}