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: 8.3755
Objective: 8.4853
Reward Accuracy: 0.6261
Logp Accuracy: 0.5450
Log Diff Policy: 2.3621
Chosen Logps: -94.1866
Rejected Logps: -96.5486
Chosen Rewards: 0.0248
Rejected Rewards: -0.0742
Logits: -2.1694

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
9.358	0.1517	100	9.4767	9.4748	0.5584	0.5282	0.9202	-92.3263	-93.2465	0.1178	0.0909	-2.1941
8.4865	0.3033	200	8.7651	8.8251	0.6121	0.5436	1.7419	-92.0866	-93.8284	0.1298	0.0618	-2.2153
8.2077	0.4550	300	8.5507	8.6214	0.6264	0.5461	2.2164	-92.5313	-94.7477	0.1076	0.0158	-2.2241
7.7431	0.6067	400	8.4722	8.5772	0.6314	0.5503	2.3954	-90.6295	-93.0249	0.2026	0.1020	-2.2473
7.822	0.7583	500	8.4631	8.5834	0.6284	0.5445	2.3357	-89.8860	-92.2217	0.2398	0.1421	-2.2090
7.6491	0.9100	600	8.4053	8.5433	0.6295	0.5445	2.2425	-91.4902	-93.7327	0.1596	0.0666	-2.2060

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-IPO-25-2

mistral-7b-expo-7b-IPO-25-2

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

hZzy/
mistral-7b-expo-7b-IPO-25-2