xtremedistil-l12-h384-uncased-zeroshot-v1.1-none
A slightly larger sibling to https://hf.co/MoritzLaurer/xtremedistil-l6-h256-zeroshot-v1.1-all-33
Model description
This model is a fine-tuned version of microsoft/xtremedistil-l12-h384-uncased on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.2063
- F1 Macro: 0.5570
- F1 Micro: 0.6385
- Accuracy Balanced: 0.6104
- Accuracy: 0.6385
- Precision Macro: 0.5705
- Recall Macro: 0.6104
- Precision Micro: 0.6385
- Recall Micro: 0.6385
Training and evaluation data
See https://github.com/MoritzLaurer/zeroshot-classifier/blob/main/datasets_overview.csv
Training procedure
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 5e-05
- train_batch_size: 32
- eval_batch_size: 32
- seed: 80085
- gradient_accumulation_steps: 2
- total_train_batch_size: 64
- optimizer: Adam with betas=(0.9,0.98) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_ratio: 0.04
- num_epochs: 3
Training results
Training Loss | Epoch | Step | Validation Loss | F1 Macro | F1 Micro | Accuracy Balanced | Accuracy | Precision Macro | Recall Macro | Precision Micro | Recall Micro |
---|---|---|---|---|---|---|---|---|---|---|---|
0.2756 | 0.32 | 5000 | 0.4155 | 0.8146 | 0.8255 | 0.8215 | 0.8255 | 0.8101 | 0.8215 | 0.8255 | 0.8255 |
0.2395 | 0.65 | 10000 | 0.4166 | 0.8182 | 0.8303 | 0.8222 | 0.8303 | 0.8151 | 0.8222 | 0.8303 | 0.8303 |
0.2464 | 0.97 | 15000 | 0.4114 | 0.8204 | 0.8325 | 0.8239 | 0.8325 | 0.8175 | 0.8239 | 0.8325 | 0.8325 |
0.2105 | 1.3 | 20000 | 0.4051 | 0.8236 | 0.8363 | 0.8254 | 0.8363 | 0.8219 | 0.8254 | 0.8363 | 0.8363 |
0.2267 | 1.62 | 25000 | 0.4030 | 0.8244 | 0.8373 | 0.8257 | 0.8373 | 0.8231 | 0.8257 | 0.8373 | 0.8373 |
0.2312 | 1.95 | 30000 | 0.4088 | 0.8233 | 0.836 | 0.8250 | 0.836 | 0.8217 | 0.8250 | 0.836 | 0.836 |
0.2241 | 2.27 | 35000 | 0.4061 | 0.8257 | 0.8375 | 0.8291 | 0.8375 | 0.8229 | 0.8291 | 0.8375 | 0.8375 |
0.2183 | 2.6 | 40000 | 0.4043 | 0.8259 | 0.838 | 0.8285 | 0.838 | 0.8235 | 0.8285 | 0.838 | 0.838 |
0.2285 | 2.92 | 45000 | 0.4041 | 0.8241 | 0.8365 | 0.8263 | 0.8365 | 0.8220 | 0.8263 | 0.8365 | 0.8365 |
Framework versions
- Transformers 4.36.2
- Pytorch 2.1.2+cu121
- Datasets 2.16.1
- Tokenizers 0.15.0
- Downloads last month
- 11
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social
visibility and check back later, or deploy to Inference Endpoints (dedicated)
instead.
Model tree for pszemraj/xtremedistil-l12-h384-uncased-zeroshot-v1.1
Base model
microsoft/xtremedistil-l12-h384-uncased
Finetuned
this model