This model is a fine-tuned version of trl-lib/qwen1.5-0.5b-sft on the yakazimir/ultrafeedback_binarized dataset. It achieves the following results on the evaluation set:

Loss: 0.0000
Rewards/chosen: -45.0891
Rewards/rejected: -46.1094
Rewards/accuracies: 0.5682
Rewards/margins: 1.0204
Logps/rejected: -46.1094
Logps/chosen: -45.0891
Logits/rejected: 7.4245
Logits/chosen: 7.7499

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: 1e-06
train_batch_size: 2
eval_batch_size: 4
seed: 42
distributed_type: multi-GPU
gradient_accumulation_steps: 16
total_train_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: 3.0

Training results

Training Loss	Epoch	Step	Validation Loss	Rewards/chosen	Rewards/rejected	Rewards/accuracies	Rewards/margins	Logps/rejected	Logps/chosen	Logits/rejected	Logits/chosen
0.0	0.2141	400	0.0001	-30.9341	-32.7194	0.5697	1.7853	-32.7194	-30.9341	4.5291	4.5264
0.0	0.4282	800	0.0000	-38.7534	-40.2065	0.5593	1.4531	-40.2065	-38.7534	6.2341	6.3877
0.0009	0.6422	1200	0.0000	-38.5460	-39.9578	0.5512	1.4119	-39.9578	-38.5460	6.1244	6.2779
0.0	0.8563	1600	0.0000	-40.0222	-41.4115	0.5690	1.3893	-41.4115	-40.0222	6.5494	6.7346
0.0	1.0704	2000	0.0000	-43.0566	-44.2275	0.5653	1.1709	-44.2275	-43.0566	7.0818	7.3504
0.0	1.2845	2400	0.0000	-43.5288	-44.6477	0.5645	1.1189	-44.6477	-43.5288	7.1882	7.4775
0.0	1.4986	2800	0.0000	-43.7383	-44.8584	0.5660	1.1201	-44.8584	-43.7383	7.1745	7.4634
0.0	1.7127	3200	0.0000	-44.4950	-45.5556	0.5638	1.0605	-45.5556	-44.4950	7.2848	7.5950
0.0	1.9267	3600	0.0000	-44.5958	-45.6569	0.5645	1.0611	-45.6569	-44.5958	7.2814	7.5948
0.0	2.1408	4000	0.0000	-44.8271	-45.8411	0.5668	1.0140	-45.8411	-44.8271	7.4235	7.7436
0.0	2.3549	4400	0.0000	-45.1344	-46.1374	0.5653	1.0030	-46.1374	-45.1344	7.3526	7.6831
0.0	2.5690	4800	0.0000	-45.0201	-46.0501	0.5653	1.0300	-46.0501	-45.0201	7.3843	7.7103
0.0	2.7831	5200	0.0000	-45.3432	-46.3394	0.5653	0.9961	-46.3394	-45.3432	7.4499	7.7830
0.0	2.9972	5600	0.0000	-45.0891	-46.1094	0.5682	1.0204	-46.1094	-45.0891	7.4245	7.7499

Framework versions

Transformers 4.44.2
Pytorch 2.2.2+cu121
Datasets 2.18.0
Tokenizers 0.19.1

RichardErkhov/yakazimir_-_qwen_cfUNL_entropy-awq

Original model description:

qwen_cfUNL_entropy

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

RichardErkhov/
yakazimir_-_qwen_cfUNL_entropy-awq