sft_uniqueness_new

This model is a fine-tuned version of Qwen/Qwen2-VL-7B-Instruct on the enrico, the fer2013, the resisc45, the decimer, the ucmerced and the inaturalist datasets. It achieves the following results on the evaluation set:

  • Loss: 0.2041

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: 4
  • total_train_batch_size: 8
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: cosine
  • lr_scheduler_warmup_ratio: 0.1
  • num_epochs: 3.0

Training results

Training Loss Epoch Step Validation Loss
0.736 0.0454 500 0.6253
0.4861 0.0908 1000 0.3907
0.4181 0.1362 1500 0.3503
0.3028 0.1817 2000 0.3344
0.3214 0.2271 2500 0.3016
0.303 0.2725 3000 0.2977
0.3694 0.3179 3500 0.2994
0.3416 0.3633 4000 0.2988
0.266 0.4087 4500 0.2733
0.2433 0.4542 5000 0.2824
0.2216 0.4996 5500 0.2608
0.2781 0.5450 6000 0.2595
0.2206 0.5904 6500 0.2495
0.2403 0.6358 7000 0.2441
0.2681 0.6812 7500 0.2487
0.2041 0.7266 8000 0.2309
0.2982 0.7721 8500 0.2371
0.2233 0.8175 9000 0.2332
0.2416 0.8629 9500 0.2305
0.1913 0.9083 10000 0.2288
0.2006 0.9537 10500 0.2316
0.1846 0.9991 11000 0.2236
0.2535 1.0446 11500 0.2257
0.1195 1.0900 12000 0.2257
0.1386 1.1354 12500 0.2197
0.1542 1.1808 13000 0.2315
0.1951 1.2262 13500 0.2194
0.1833 1.2716 14000 0.2194
0.1244 1.3170 14500 0.2179
0.1624 1.3625 15000 0.2153
0.2119 1.4079 15500 0.2152
0.1696 1.4533 16000 0.2227
0.1398 1.4987 16500 0.2123
0.2048 1.5441 17000 0.2136
0.1115 1.5895 17500 0.2082
0.2041 1.6350 18000 0.2004
0.2027 1.6804 18500 0.1996
0.1198 1.7258 19000 0.2000
0.1837 1.7712 19500 0.2014
0.1748 1.8166 20000 0.1982
0.156 1.8620 20500 0.1981
0.1704 1.9074 21000 0.1924
0.1532 1.9529 21500 0.1963
0.1719 1.9983 22000 0.1920
0.0699 2.0437 22500 0.2018
0.145 2.0891 23000 0.2079
0.1097 2.1345 23500 0.2018
0.1007 2.1799 24000 0.2035
0.0622 2.2254 24500 0.2074
0.095 2.2708 25000 0.2000
0.144 2.3162 25500 0.2056
0.2398 2.3616 26000 0.2032
0.0303 2.4070 26500 0.2016
0.0766 2.4524 27000 0.2044
0.0822 2.4978 27500 0.2029
0.1465 2.5433 28000 0.2057
0.094 2.5887 28500 0.2006
0.1033 2.6341 29000 0.2012
0.128 2.6795 29500 0.2027
0.0784 2.7249 30000 0.2035
0.1244 2.7703 30500 0.2045
0.1106 2.8158 31000 0.2042
0.0845 2.8612 31500 0.2042
0.1129 2.9066 32000 0.2041
0.1064 2.9520 32500 0.2041
0.1087 2.9974 33000 0.2041

Framework versions

  • PEFT 0.12.0
  • Transformers 4.45.2
  • Pytorch 2.1.2+cu121
  • Datasets 3.1.0
  • Tokenizers 0.20.3
Downloads last month
-
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support

Model tree for sxj1215/Qwen2-VL-Uniqueness

Base model

Qwen/Qwen2-VL-7B
Adapter
(171)
this model

Collection including sxj1215/Qwen2-VL-Uniqueness