Hugging Face
Models
Datasets
Spaces
Posts
Docs
Solutions
Pricing
Log In
Sign Up
dball
/
zephyr-7b-dpo-qlora-no-sft
like
0
PEFT
TensorBoard
Safetensors
HuggingFaceH4/ultrafeedback_binarized
mistral
alignment-handbook
Generated from Trainer
trl
dpo
4-bit precision
bitsandbytes
License:
apache-2.0
Model card
Files
Files and versions
Metrics
Training metrics
Community
1
Train
Use this model
571698d
zephyr-7b-dpo-qlora-no-sft
/
train_results.json
dball
Model save
42d7b6a
verified
5 months ago
raw
history
blame
No virus
196 Bytes
{
"epoch"
:
1.0
,
"train_loss"
:
0.5756704107174588
,
"train_runtime"
:
163576.2451
,
"train_samples"
:
61135
,
"train_samples_per_second"
:
0.374
,
"train_steps_per_second"
:
0.047
}