algsteer / README.md
yilunhao's picture
Fine-tuned model
c4ba3bf verified
metadata
library_name: peft
license: apache-2.0
base_model: Qwen/Qwen3-8B
tags:
  - base_model:adapter:Qwen/Qwen3-8B
  - lora
  - transformers
pipeline_tag: text-generation
model-index:
  - name: algsteer
    results: []

algsteer

This model is a fine-tuned version of Qwen/Qwen3-8B on an unknown dataset. It achieves the following results on the evaluation set:

  • Loss: 0.1092

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: 0.0003
  • train_batch_size: 1
  • eval_batch_size: 1
  • seed: 42
  • distributed_type: multi-GPU
  • num_devices: 8
  • gradient_accumulation_steps: 16
  • total_train_batch_size: 128
  • total_eval_batch_size: 8
  • optimizer: Use adamw_torch_fused with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
  • lr_scheduler_type: linear
  • lr_scheduler_warmup_steps: 100
  • num_epochs: 50
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss
0.3399 1.3909 50 0.0722
0.0379 2.7818 100 0.0320
0.0291 4.1675 150 0.0308
0.0303 5.5585 200 0.0310
0.0265 6.9494 250 0.0306
0.0245 8.3351 300 0.0326
0.023 9.7260 350 0.0337
0.0218 11.1117 400 0.0369
0.0209 12.5026 450 0.0369
0.0202 13.8935 500 0.0388
0.0197 15.2792 550 0.0411
0.019 16.6702 600 0.0418
0.0186 18.0558 650 0.0442
0.0178 19.4468 700 0.0454
0.0174 20.8377 750 0.0473
0.0167 22.2234 800 0.0499
0.0161 23.6143 850 0.0492
0.0155 25.0 900 0.0508
0.0145 26.3909 950 0.0564
0.0138 27.7818 1000 0.0566
0.0131 29.1675 1050 0.0621
0.0122 30.5585 1100 0.0633
0.0115 31.9494 1150 0.0675
0.0105 33.3351 1200 0.0736
0.0098 34.7260 1250 0.0740
0.009 36.1117 1300 0.0807
0.0081 37.5026 1350 0.0832
0.0076 38.8935 1400 0.0855
0.007 40.2792 1450 0.0911
0.0064 41.6702 1500 0.0946
0.0061 43.0558 1550 0.0972
0.0056 44.4468 1600 0.1019
0.0055 45.8377 1650 0.1036
0.0053 47.2234 1700 0.1064
0.0051 48.6143 1750 0.1084
0.005 50.0 1800 0.1092

Framework versions

  • PEFT 0.18.0.rc0
  • Transformers 4.57.1
  • Pytorch 2.9.0+cu128
  • Datasets 4.3.0
  • Tokenizers 0.22.2