mistral-7b-expo-7b-L2EXPO-25-final-1

hZzy/mistral-7b-expo-7b-L2EXPO-25-final-1

mistral-7b-expo-7b-L2EXPO-25-final-1 is a machine learning model from hZzy. Use it for the machine learning task on the model card, and read the license before you ship it in a product. It is set up for peft. The card lists the license as apache-2.0.

should probably proofread and complete it, then remove this comment. --

Downloads · 30 days

6% of all-time downloads

All-time downloads

Public

Repo size

465 MB

Likes

Public

Hugging Face

Repo makeup

Click a slice to open those files.

.safetensors27.3 MB · 91%

At a glance

Library: peft
License: apache-2.0
Model type: mistral
Access: Public
Created: Apr 16, 2025
Updated: Apr 17, 2025
SHA: 78e38ec0

Base models

hZzy/mistral-7b-sft-25-1

Datasets

hZzy/direction_right2

Model card

From the Hugging Face model README

mistral-7b-expo-7b-L2EXPO-25-final-1

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: 0.4434
Objective: 0.4446
Reward Accuracy: 0.6686
Logp Accuracy: 0.6605
Log Diff Policy: 20.4754
Chosen Logps: -191.0983
Rejected Logps: -211.5737
Chosen Rewards: -0.9642
Rejected Rewards: -1.1651
Logits: -2.0799

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: 0.2
num_epochs: 2

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
0.617	0.1213	80	0.5119	0.5088	0.5414	0.5165	0.4249	-93.8795	-94.3044	0.0080	0.0076	-2.2005
0.5757	0.2427	160	0.5083	0.5054	0.5660	0.5254	0.8705	-90.6804	-91.5509	0.0400	0.0351	-2.1666
0.5439	0.3640	240	0.4913	0.4904	0.5842	0.5671	4.1006	-113.6787	-117.7794	-0.1900	-0.2271	-1.9351
0.5254	0.4853	320	0.4776	0.4800	0.6119	0.6026	9.7092	-145.8312	-155.5405	-0.5115	-0.6048	-1.9366
0.4916	0.6067	400	0.4666	0.4651	0.6415	0.6300	12.8018	-132.8940	-145.6958	-0.3821	-0.5063	-2.0733
0.4908	0.7280	480	0.4599	0.4614	0.6477	0.6381	15.0878	-167.3499	-182.4376	-0.7267	-0.8737	-2.0208
0.4998	0.8493	560	0.4461	0.4463	0.6591	0.6468	13.4053	-149.0239	-162.4292	-0.5434	-0.6736	-1.9914
0.4547	0.9707	640	0.4524	0.4497	0.6681	0.6549	17.6039	-173.2037	-190.8076	-0.7852	-0.9574	-2.0256
0.4464	1.0920	720	0.4526	0.4508	0.6695	0.6569	18.8492	-145.8253	-164.6745	-0.5114	-0.6961	-2.0653
0.4678	1.2133	800	0.4501	0.4486	0.6697	0.6577	18.9049	-169.4865	-188.3915	-0.7480	-0.9333	-2.0396
0.4597	1.3347	880	0.4440	0.4423	0.6731	0.6527	16.6871	-181.3906	-198.0777	-0.8671	-1.0301	-2.0012
0.4762	1.4560	960	0.4436	0.4418	0.6669	0.6591	18.4037	-146.4813	-164.8850	-0.5180	-0.6982	-1.9367
0.4297	1.5774	1040	0.4389	0.4381	0.6703	0.6538	17.7650	-166.6099	-184.3749	-0.7193	-0.8931	-2.1020
0.4207	1.6987	1120	0.4413	0.4421	0.6756	0.6602	19.4871	-190.2710	-209.7581	-0.9559	-1.1469	-2.0698
0.4131	1.8200	1200	0.4401	0.4428	0.6728	0.6600	20.1653	-190.6565	-210.8218	-0.9597	-1.1576	-2.0651
0.4035	1.9414	1280	0.4402	0.4408	0.6686	0.6532	18.6035	-190.2284	-208.8319	-0.9555	-1.1377	-2.0701

Framework versions

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

Library: peft
Type: mistral
License: apache-2.0
Created: Apr 16, 2025
Updated: Apr 17, 2025