metadata

license: mit
base_model: TheBloke/zephyr-7B-beta-GPTQ
tags:
  - trl
  - sft
  - generated_from_trainer
metrics:
  - rouge
model-index:
  - name: zephyr-support-chatbot
    results: []

zephyr-support-chatbot

This model is a fine-tuned version of TheBloke/zephyr-7B-beta-GPTQ on the None dataset. It achieves the following results on the evaluation set:

Loss: 1.2805
Rouge1: 0.6842
Rouge2: 0.4855
Rougel: 0.6563
Rougelsum: 0.6711

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: 2e-05
train_batch_size: 16
eval_batch_size: 8
seed: 42
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: cosine
num_epochs: 20
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Rouge1	Rouge2	Rougel	Rougelsum
2.422	1.11	10	2.7640	0.4291	0.1054	0.3461	0.3890
2.2454	2.22	20	2.5777	0.4423	0.1184	0.3607	0.4034
2.1454	3.33	30	2.3809	0.4713	0.1437	0.3860	0.4288
1.9437	4.44	40	2.1804	0.5021	0.1646	0.4027	0.4598
1.7975	5.56	50	2.0124	0.5355	0.1786	0.4425	0.4941
1.6621	6.67	60	1.8249	0.5540	0.2188	0.5011	0.5348
1.5141	7.78	70	1.6004	0.6161	0.3377	0.5701	0.5961
1.3291	8.89	80	1.4718	0.6513	0.3903	0.6072	0.6322
1.2206	10.0	90	1.3916	0.6652	0.4218	0.6265	0.6471
1.1767	11.11	100	1.3339	0.6840	0.4769	0.6489	0.6675
1.1462	12.22	110	1.3115	0.6807	0.4785	0.6506	0.6665
1.0924	13.33	120	1.2993	0.6843	0.4842	0.6539	0.6701
1.0602	14.44	130	1.2917	0.6854	0.4845	0.6561	0.6717
1.1177	15.56	140	1.2863	0.6835	0.4842	0.6547	0.6703
1.0756	16.67	150	1.2830	0.6838	0.4825	0.6549	0.6705
1.0894	17.78	160	1.2813	0.6838	0.4844	0.6560	0.6719
1.0649	18.89	170	1.2806	0.6842	0.4855	0.6563	0.6711
1.1019	20.0	180	1.2805	0.6842	0.4855	0.6563	0.6711

Framework versions

Transformers 4.35.2
Pytorch 2.1.0+cu121
Datasets 2.16.0
Tokenizers 0.15.0