This model is a fine-tuned version of hZzy/mistral-7b-sft-25-1 on the hZzy/direction_right2 dataset. It achieves the following results on the evaluation set:

Loss: 0.6090
Objective: 0.6216
Reward Accuracy: 0.6622
Logp Accuracy: 0.6317
Log Diff Policy: 9.5533
Chosen Logps: -124.4338
Rejected Logps: -133.9871
Chosen Rewards: -1.4876
Rejected Rewards: -1.9461
Logits: -2.1433

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: 5e-06
train_batch_size: 3
eval_batch_size: 3
seed: 42
distributed_type: multi-GPU
num_devices: 3
gradient_accumulation_steps: 12
total_train_batch_size: 108
total_eval_batch_size: 9
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: constant_with_warmup
lr_scheduler_warmup_ratio: 0.1
num_epochs: 1

Training results

Training Loss	Epoch	Step	Validation Loss	Objective	Reward Accuracy	Logp Accuracy	Log Diff Policy	Chosen Logps	Rejected Logps	Chosen Rewards	Rejected Rewards	Logits
0.6912	0.0758	50	0.6916	0.6916	0.5453	0.5165	0.4536	-93.0364	-93.4900	0.0823	0.0787	-2.1999
0.6816	0.1517	100	0.6839	0.6835	0.5682	0.5268	0.8519	-92.7664	-93.6183	0.0958	0.0723	-2.2118
0.6682	0.2275	150	0.6681	0.6699	0.5898	0.5391	1.8524	-90.8783	-92.7307	0.1902	0.1167	-2.1819
0.6368	0.3033	200	0.6511	0.6563	0.6099	0.5663	3.4226	-95.6494	-99.0720	-0.0484	-0.2004	-2.1272
0.6147	0.3792	250	0.6438	0.6524	0.6230	0.5800	4.8890	-106.6454	-111.5345	-0.5982	-0.8235	-2.1329
0.6198	0.4550	300	0.6340	0.6433	0.6323	0.5979	5.6663	-101.5432	-107.2095	-0.3430	-0.6073	-2.1344
0.5733	0.5308	350	0.6317	0.6468	0.6401	0.6119	7.1300	-111.3464	-118.4764	-0.8332	-1.1706	-2.1173
0.5624	0.6067	400	0.6249	0.6382	0.6477	0.6099	7.5147	-99.6262	-107.1409	-0.2472	-0.6038	-2.1473
0.5804	0.6825	450	0.6201	0.6366	0.6490	0.6270	8.6628	-115.1139	-123.7767	-1.0216	-1.4356	-2.1786
0.5731	0.7583	500	0.6207	0.6348	0.6544	0.6326	9.4271	-112.7352	-122.1623	-0.9026	-1.3549	-2.1662
0.5566	0.8342	550	0.6190	0.6309	0.6549	0.6342	9.5743	-120.8242	-130.3985	-1.3071	-1.7667	-2.1212
0.5574	0.9100	600	0.6111	0.6232	0.6647	0.6367	9.6758	-128.2970	-137.9727	-1.6807	-2.1454	-2.1279
0.5855	0.9858	650	0.6065	0.6193	0.6672	0.6342	9.7652	-122.9088	-132.6740	-1.4113	-1.8805	-2.1149

Framework versions

PEFT 0.11.1
Transformers 4.42.0
Pytorch 2.6.0+cu124
Datasets 3.2.0
Tokenizers 0.19.1

hZzy/mistral-7b-expo-7b-DPO-25-last-2

mistral-7b-expo-7b-DPO-25-last-2

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

hZzy/
mistral-7b-expo-7b-DPO-25-last-2