mistralit2_1000_STEPS_1e8_rate_05_beta_DPO

tsavage68/mistralit2_1000_STEPS_1e8_rate_05_beta_DPO

mistralit2_1000_STEPS_1e8_rate_05_beta_DPO is a text generation model from tsavage68. Use it when you need the model to write or continue text. It is set up for transformers. The card lists the license as apache-2.0.

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

Downloads · 30 days

9% of all-time downloads

All-time downloads

Public

Parameters

7.2B

14.5 GB on disk

Likes

Public

Hugging Face

Repo makeup

Click a slice to open those files.

.safetensors29 GB · 100%

At a glance

Task: Text Generation
Library: transformers
License: apache-2.0
Model type: mistral
Access: Public
Created: Mar 11, 2024
Updated: Mar 11, 2024
SHA: cc35b3e2

Try a prompt

Base models

mistralai/Mistral-7B-Instruct-v0.2

Model card

From the Hugging Face model README

mistralit2_1000_STEPS_1e8_rate_05_beta_DPO

This model is a fine-tuned version of mistralai/Mistral-7B-Instruct-v0.2 on an unknown dataset. It achieves the following results on the evaluation set:

Loss: 0.6855
Rewards/chosen: -0.0282
Rewards/rejected: -0.0460
Rewards/accuracies: 0.5253
Rewards/margins: 0.0178
Logps/rejected: -28.6643
Logps/chosen: -23.4421
Logits/rejected: -2.8649
Logits/chosen: -2.8652

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-08
train_batch_size: 4
eval_batch_size: 1
seed: 42
gradient_accumulation_steps: 2
total_train_batch_size: 8
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: cosine
lr_scheduler_warmup_steps: 100
training_steps: 1000

Training results

Training Loss	Epoch	Step	Validation Loss	Rewards/chosen	Rewards/rejected	Rewards/accuracies	Rewards/margins	Logps/rejected	Logps/chosen	Logits/rejected	Logits/chosen
0.6906	0.1	50	0.6922	0.0037	-0.0004	0.4549	0.0041	-28.5732	-23.3783	-2.8653	-2.8655
0.6932	0.2	100	0.6921	-0.0005	-0.0047	0.4593	0.0042	-28.5818	-23.3868	-2.8653	-2.8655
0.6966	0.29	150	0.6897	-0.0096	-0.0189	0.4593	0.0093	-28.6102	-23.4050	-2.8650	-2.8653
0.6871	0.39	200	0.6884	-0.0191	-0.0309	0.5143	0.0119	-28.6343	-23.4239	-2.8651	-2.8654
0.6964	0.49	250	0.6899	-0.0092	-0.0180	0.4989	0.0089	-28.6085	-23.4042	-2.8650	-2.8652
0.69	0.59	300	0.6860	-0.0114	-0.0279	0.5473	0.0165	-28.6282	-23.4086	-2.8650	-2.8653
0.6912	0.68	350	0.6901	-0.0242	-0.0329	0.4703	0.0087	-28.6382	-23.4342	-2.8650	-2.8653
0.6914	0.78	400	0.6893	-0.0279	-0.0380	0.5121	0.0102	-28.6485	-23.4415	-2.8649	-2.8652
0.6874	0.88	450	0.6885	-0.0222	-0.0339	0.4945	0.0116	-28.6401	-23.4303	-2.8651	-2.8653
0.6858	0.98	500	0.6893	-0.0299	-0.0402	0.5077	0.0102	-28.6527	-23.4457	-2.8650	-2.8652
0.6849	1.07	550	0.6883	-0.0300	-0.0423	0.4967	0.0123	-28.6570	-23.4459	-2.8650	-2.8653
0.6896	1.17	600	0.6896	-0.0287	-0.0381	0.5143	0.0093	-28.6485	-23.4433	-2.8649	-2.8652
0.6883	1.27	650	0.6882	-0.0258	-0.0384	0.4879	0.0126	-28.6491	-23.4374	-2.8650	-2.8653
0.6932	1.37	700	0.6883	-0.0325	-0.0448	0.4989	0.0123	-28.6621	-23.4509	-2.8649	-2.8651
0.6863	1.46	750	0.6899	-0.0283	-0.0372	0.5033	0.0089	-28.6468	-23.4424	-2.8650	-2.8653
0.693	1.56	800	0.6922	-0.0345	-0.0387	0.4813	0.0042	-28.6498	-23.4548	-2.8650	-2.8652
0.6858	1.66	850	0.6849	-0.0275	-0.0465	0.5297	0.0190	-28.6654	-23.4408	-2.8649	-2.8652
0.6924	1.76	900	0.6855	-0.0282	-0.0460	0.5253	0.0178	-28.6643	-23.4421	-2.8649	-2.8652
0.7048	1.86	950	0.6855	-0.0282	-0.0460	0.5253	0.0178	-28.6643	-23.4421	-2.8649	-2.8652
0.6835	1.95	1000	0.6855	-0.0282	-0.0460	0.5253	0.0178	-28.6643	-23.4421	-2.8649	-2.8652

Framework versions

Transformers 4.38.2
Pytorch 2.0.0+cu117
Datasets 2.18.0
Tokenizers 0.15.2

Task: Text Generation
Library: transformers
Type: mistral
License: apache-2.0
Created: Mar 11, 2024
Updated: Mar 11, 2024