0428

This model is a fine-tuned version of ../../models/Qwen1.5-7B-sft-0425 on the alpaca_formatted_review_new_data_greater_7 dataset. It achieves the following results on the evaluation set:

Loss: 1.0733

Model description

Qwen1.5 is the beta version of Qwen2, a transformer-based decoder-only language model pretrained on a large amount of data. In comparison with the previous released Qwen, the improvements include:

8 model sizes, including 0.5B, 1.8B, 4B, 7B, 14B, 32B and 72B dense models, and an MoE model of 14B with 2.7B activated;
Significant performance improvement in Chat models;
Multilingual support of both base and chat models;
Stable support of 32K context length for models of all sizes
No need of trust_remote_code.

For more details, please refer to the blog post and GitHub repo.

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-05
train_batch_size: 2
eval_batch_size: 1
seed: 42
distributed_type: multi-GPU
num_devices: 2
gradient_accumulation_steps: 2
total_train_batch_size: 8
total_eval_batch_size: 2
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: cosine
lr_scheduler_warmup_steps: 5
num_epochs: 5.0
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
0.8554	0.25	10	1.1541
0.6139	0.5	20	1.1258
0.629	0.75	30	1.1057
0.7943	1.0	40	1.0993
0.6658	1.25	50	1.0964
0.778	1.5	60	1.0892
0.593	1.75	70	1.0868
0.8847	2.0	80	1.0816
0.5067	2.25	90	1.0806
0.9706	2.5	100	1.0789
0.7302	2.75	110	1.0763
0.6855	3.0	120	1.0768
0.4358	3.25	130	1.0754
0.5777	3.5	140	1.0740
0.5687	3.75	150	1.0732
0.6462	4.0	160	1.0732
0.5465	4.25	170	1.0733
0.7926	4.5	180	1.0737
0.4968	4.75	190	1.0735
0.6406	5.0	200	1.0733

Framework versions

PEFT 0.10.0
Transformers 4.40.0
Pytorch 2.1.0+cu121
Datasets 2.14.5
Tokenizers 0.19.1