metadata

base_model: Qwen/Qwen2-72B-Instruct
library_name: peft
license: other
tags:
  - llama-factory
  - lora
  - generated_from_trainer
model-index:
  - name: Qwen2-72B-Instruct
    results: []

Qwen2-72B-Instruct

This model is a fine-tuned version of Qwen/Qwen2-72B-Instruct on the alpaca_mac dataset. It achieves the following results on the evaluation set:

Loss: 2.6303

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: 0.0001
train_batch_size: 2
eval_batch_size: 1
seed: 42
gradient_accumulation_steps: 8
total_train_batch_size: 16
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: cosine
lr_scheduler_warmup_ratio: 0.1
num_epochs: 6.0

Training results

Training Loss	Epoch	Step	Validation Loss
1.5022	0.2499	70	1.4518
1.4121	0.4998	140	1.3727
1.4038	0.7497	210	1.3051
1.2739	0.9996	280	1.2890
1.1436	1.2494	350	1.3195
1.0783	1.4993	420	1.3106
1.1219	1.7492	490	1.3045
1.0966	1.9991	560	1.3094
0.6134	2.2490	630	1.4946
0.6342	2.4989	700	1.4859
0.6665	2.7488	770	1.5236
0.6101	2.9987	840	1.5220
0.2467	3.2485	910	1.8390
0.2284	3.4984	980	1.8253
0.2839	3.7483	1050	1.8688
0.2111	3.9982	1120	1.8910
0.0753	4.2481	1190	2.2224
0.072	4.4980	1260	2.3093
0.0351	4.7479	1330	2.2221
0.0644	4.9978	1400	2.2804
0.0257	5.2477	1470	2.5593
0.0249	5.4975	1540	2.6220
0.0238	5.7474	1610	2.6189
0.0262	5.9973	1680	2.6303

Framework versions

PEFT 0.11.1
Transformers 4.43.3
Pytorch 2.4.0+cu121
Datasets 2.19.1
Tokenizers 0.19.1