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: 0.4398
Objective: 0.4395
Reward Accuracy: 0.6773
Logp Accuracy: 0.6616
Log Diff Policy: 16.3542
Chosen Logps: -155.5914
Rejected Logps: -171.9456
Chosen Rewards: -0.6091
Rejected Rewards: -0.7688
Logits: -2.1212

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: 0.15
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
0.587	0.0758	50	0.5118	0.5088	0.5274	0.5162	0.4235	-93.6864	-94.1099	0.0100	0.0095	-2.1994
0.6103	0.1517	100	0.5091	0.5061	0.5492	0.5204	0.7358	-92.5364	-93.2722	0.0215	0.0179	-2.1892
0.6304	0.2275	150	0.4989	0.4965	0.5749	0.5456	2.2257	-89.6398	-91.8654	0.0504	0.0320	-2.1138
0.5416	0.3033	200	0.4898	0.4922	0.5973	0.5889	7.3559	-121.1515	-128.5074	-0.2647	-0.3344	-1.8905
0.5415	0.3792	250	0.4804	0.4799	0.6068	0.6004	9.1753	-116.7436	-125.9189	-0.2206	-0.3085	-1.8266
0.535	0.4550	300	0.4746	0.4744	0.6292	0.6267	12.9555	-134.9998	-147.9553	-0.4032	-0.5289	-1.9479
0.5001	0.5308	350	0.4657	0.4666	0.6326	0.6292	13.2013	-146.9910	-160.1923	-0.5231	-0.6513	-2.0229
0.4802	0.6067	400	0.4611	0.4610	0.6418	0.6353	13.9345	-130.1516	-144.0861	-0.3547	-0.4902	-2.1343
0.477	0.6825	450	0.4540	0.4551	0.6560	0.6488	15.1873	-153.7426	-168.9298	-0.5906	-0.7387	-2.0982
0.4996	0.7583	500	0.4532	0.4544	0.6594	0.6605	16.8046	-154.2913	-171.0958	-0.5961	-0.7603	-2.2076
0.4394	0.8342	550	0.4605	0.4635	0.6572	0.6504	18.2414	-168.7838	-187.0253	-0.7410	-0.9196	-2.2567
0.4646	0.9100	600	0.4501	0.4515	0.6619	0.6614	17.1737	-181.8153	-198.9890	-0.8713	-1.0392	-2.1748
0.493	0.9858	650	0.4408	0.4404	0.6751	0.6602	16.3453	-155.9185	-172.2638	-0.6124	-0.7720	-2.1097

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-L1EXPO-25-last-try-2

mistral-7b-expo-7b-L1EXPO-25-last-try-2

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

hZzy/
mistral-7b-expo-7b-L1EXPO-25-last-try-2