You can deploy this model and start earning money today!

aaaalongaa/
llama3-8b-mypo3_sim-full-beta5.0-lr4e-7

Name: llama3-8b-mypo3_sim-full-beta5.0-lr4e-7
Rating: 2.5 (7 reviews)

generated_from_trainer

HuggingFaceH4/ultrafeedback_binarized

princeton-nlp/Llama-3-Base-8B-SFT

finetune

Views

No views yet

llama3-8b-mypo3_sim-full-beta5.0-lr4e-7

This model is a fine-tuned version of princeton-nlp/Llama-3-Base-8B-SFT on the HuggingFaceH4/ultrafeedback_binarized dataset. It achieves the following results on the evaluation set:

Loss: 1.3561
Rewards/chosen: 0.0455
Rewards/rejected: -0.3436
Rewards/accuracies: 0.7540
Rewards/margins: 0.3891
Logps/rejected: -1.5581
Logps/chosen: -1.2620
Logits/rejected: -1.1313
Logits/chosen: -1.1093

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 4e-07
train_batch_size: 4
eval_batch_size: 8
seed: 42
distributed_type: multi-GPU
num_devices: 4
gradient_accumulation_steps: 2
total_train_batch_size: 32
total_eval_batch_size: 32
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: cosine
lr_scheduler_warmup_ratio: 0.1
num_epochs: 1

Training results

Training Loss	Epoch	Step	Validation Loss	Rewards/chosen	Rewards/rejected	Rewards/accuracies	Rewards/margins	Logps/rejected	Logps/chosen	Logits/rejected	Logits/chosen
1.3789	0.0523	100	1.3808	-0.0274	-0.0785	0.6806	0.0511	-1.5051	-1.2766	-1.0193	-0.9896
1.3769	0.1047	200	1.3762	-0.0321	-0.1675	0.7004	0.1354	-1.5229	-1.2776	-1.0582	-1.0282
1.3697	0.1570	300	1.3709	0.0326	-0.1862	0.7421	0.2188	-1.5267	-1.2646	-1.0908	-1.0623
1.3606	0.2094	400	1.3691	0.0321	-0.2135	0.7282	0.2457	-1.5321	-1.2647	-1.0704	-1.0452
1.364	0.2617	500	1.3698	-0.0392	-0.3556	0.7401	0.3164	-1.5605	-1.2790	-1.0918	-1.0669
1.3676	0.3141	600	1.3679	0.0130	-0.3015	0.7262	0.3145	-1.5497	-1.2686	-1.1094	-1.0848
1.3496	0.3664	700	1.3782	-0.0358	-0.3804	0.7242	0.3446	-1.5655	-1.2783	-1.0986	-1.0754
1.4003	0.4187	800	1.3661	0.0061	-0.3269	0.7321	0.3330	-1.5548	-1.2699	-1.1183	-1.0952
1.3764	0.4711	900	1.3641	0.0365	-0.3162	0.7401	0.3527	-1.5526	-1.2638	-1.0986	-1.0773
1.3558	0.5234	1000	1.3631	0.0963	-0.2830	0.7440	0.3793	-1.5460	-1.2519	-1.1391	-1.1166
1.3666	0.5758	1100	1.3637	0.1255	-0.2347	0.7401	0.3602	-1.5363	-1.2460	-1.1421	-1.1197
1.3602	0.6281	1200	1.3610	-0.0028	-0.3955	0.7579	0.3927	-1.5685	-1.2717	-1.1304	-1.1082
1.3466	0.6805	1300	1.3602	0.0088	-0.3839	0.7520	0.3928	-1.5662	-1.2694	-1.1200	-1.0985
1.3473	0.7328	1400	1.3606	0.0830	-0.2992	0.7440	0.3822	-1.5492	-1.2546	-1.1331	-1.1106
1.3559	0.7851	1500	1.3578	0.0615	-0.3261	0.7460	0.3876	-1.5546	-1.2588	-1.1219	-1.1003
1.3636	0.8375	1600	1.3572	0.0557	-0.3356	0.7480	0.3914	-1.5565	-1.2600	-1.1300	-1.1080
1.3529	0.8898	1700	1.3560	0.0364	-0.3568	0.7540	0.3932	-1.5608	-1.2639	-1.1243	-1.1029
1.352	0.9422	1800	1.3564	0.0487	-0.3413	0.7560	0.3900	-1.5577	-1.2614	-1.1308	-1.1089
1.3478	0.9945	1900	1.3564	0.0430	-0.3453	0.75	0.3883	-1.5585	-1.2625	-1.1265	-1.1049

Framework versions

Transformers 4.43.1
Pytorch 2.1.2+cu121
Datasets 2.18.0
Tokenizers 0.19.1

aaaalongaa/llama3-8b-mypo3_sim-full-beta5.0-lr4e-7

llama3-8b-mypo3_sim-full-beta5.0-lr4e-7

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

aaaalongaa/
llama3-8b-mypo3_sim-full-beta5.0-lr4e-7