modelId
stringlengths
6
107
label
list
readme
stringlengths
0
56.2k
readme_len
int64
0
56.2k
anahitapld/electra-base-dbd
null
--- license: apache-2.0 ---
28
tbasic5/distilbert-base-uncased-finetuned-emotion
[ "sadness", "joy", "love", "anger", "fear", "surprise" ]
--- license: apache-2.0 tags: - generated_from_trainer datasets: - emotion metrics: - accuracy - f1 model-index: - name: distilbert-base-uncased-finetuned-emotion results: - task: name: Text Classification type: text-classification dataset: name: emotion type: emotion args: default metrics: - name: Accuracy type: accuracy value: 0.925 - name: F1 type: f1 value: 0.925022224520608 --- <!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. --> # distilbert-base-uncased-finetuned-emotion This model is a fine-tuned version of [distilbert-base-uncased](https://huggingface.co/distilbert-base-uncased) on the emotion dataset. It achieves the following results on the evaluation set: - Loss: 0.2222 - Accuracy: 0.925 - F1: 0.9250 ## 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: 2e-05 - train_batch_size: 64 - eval_batch_size: 64 - seed: 42 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 - lr_scheduler_type: linear - num_epochs: 2 ### Training results | Training Loss | Epoch | Step | Validation Loss | Accuracy | F1 | |:-------------:|:-----:|:----:|:---------------:|:--------:|:------:| | 0.8521 | 1.0 | 250 | 0.3164 | 0.907 | 0.9038 | | 0.2549 | 2.0 | 500 | 0.2222 | 0.925 | 0.9250 | ### Framework versions - Transformers 4.20.1 - Pytorch 1.11.0+cu113 - Datasets 2.3.2 - Tokenizers 0.12.1
1,803
mhaegeman/autotrain-country-recognition-1059336697
[ "Austria", "Belgium", "Denmark", "Finland", "France", "Germany", "Israel", "Italy", "Netherlands", "Norway", "Poland", "Portugal", "Saudi Arabia", "South Africa", "Spain", "Sweden", "Switzerland", "Turkey", "United Arab Emirates", "United Kingdom", "United States" ]
--- tags: autotrain language: en widget: - text: "I love AutoTrain 🤗" datasets: - mhaegeman/autotrain-data-country-recognition co2_eq_emissions: 0.02952188223491361 --- # Model Trained Using AutoTrain - Problem type: Multi-class Classification - Model ID: 1059336697 - CO2 Emissions (in grams): 0.02952188223491361 ## Validation Metrics - Loss: 0.06108148396015167 - Accuracy: 0.9879569162920872 - Macro F1: 0.9765004449554612 - Micro F1: 0.9879569162920872 - Weighted F1: 0.9879450113590053 - Macro Precision: 0.9784321161207384 - Micro Precision: 0.9879569162920872 - Weighted Precision: 0.9880404765946114 - Macro Recall: 0.9748417542427885 - Micro Recall: 0.9879569162920872 - Weighted Recall: 0.9879569162920872 ## Usage You can use cURL to access this model: ``` $ curl -X POST -H "Authorization: Bearer YOUR_API_KEY" -H "Content-Type: application/json" -d '{"inputs": "I love AutoTrain"}' https://api-inference.huggingface.co/models/mhaegeman/autotrain-country-recognition-1059336697 ``` Or Python API: ``` from transformers import AutoModelForSequenceClassification, AutoTokenizer model = AutoModelForSequenceClassification.from_pretrained("mhaegeman/autotrain-country-recognition-1059336697", use_auth_token=True) tokenizer = AutoTokenizer.from_pretrained("mhaegeman/autotrain-country-recognition-1059336697", use_auth_token=True) inputs = tokenizer("I love AutoTrain", return_tensors="pt") outputs = model(**inputs) ```
1,443
Pro0100Hy6/test_trainer
[ "LABEL_0", "LABEL_1", "LABEL_2" ]
--- license: mit tags: - generated_from_trainer metrics: - accuracy model-index: - name: test_trainer results: [] --- <!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. --> # test_trainer This model is a fine-tuned version of [cointegrated/rubert-tiny](https://huggingface.co/cointegrated/rubert-tiny) on an unknown dataset. It achieves the following results on the evaluation set: - Loss: 0.7773 - Accuracy: 0.6375 ## 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: 5e-05 - train_batch_size: 8 - eval_batch_size: 8 - seed: 42 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 - lr_scheduler_type: linear - num_epochs: 1 ### Training results | Training Loss | Epoch | Step | Validation Loss | Accuracy | |:-------------:|:-----:|:----:|:---------------:|:--------:| | 0.7753 | 1.0 | 400 | 0.7773 | 0.6375 | ### Framework versions - Transformers 4.18.0 - Pytorch 1.11.0+cu113 - Datasets 2.3.2 - Tokenizers 0.12.1
1,323
Vinz9899/dumy-model
null
Entry not found
15
PGT/old_pretrained-transformer-20epochs
[ "LABEL_0", "LABEL_1", "LABEL_2", "LABEL_3", "LABEL_4", "LABEL_5", "LABEL_6" ]
Entry not found
15
dee4hf/autotrain-deephate2-1093539673
[ "Geopolitical", "Personal", "Political", "Religious" ]
--- tags: autotrain language: bn widget: - text: "I love AutoTrain 🤗" datasets: - dee4hf/autotrain-data-deephate2 co2_eq_emissions: 7.663051290039914 --- # Model Trained Using AutoTrain - Problem type: Multi-class Classification - Model ID: 1093539673 - CO2 Emissions (in grams): 7.663051290039914 ## Validation Metrics - Loss: 0.34404119849205017 - Accuracy: 0.8843120070113936 - Macro F1: 0.8771237753798016 - Micro F1: 0.8843120070113936 - Weighted F1: 0.8843498914288083 - Macro Precision: 0.8745249813256932 - Micro Precision: 0.8843120070113936 - Weighted Precision: 0.8854719661321065 - Macro Recall: 0.8812563739901838 - Micro Recall: 0.8843120070113936 - Weighted Recall: 0.8843120070113936 ## Usage You can use cURL to access this model: ``` $ curl -X POST -H "Authorization: Bearer YOUR_API_KEY" -H "Content-Type: application/json" -d '{"inputs": "I love AutoTrain"}' https://api-inference.huggingface.co/models/dee4hf/autotrain-deephate2-1093539673 ``` Or Python API: ``` from transformers import AutoModelForSequenceClassification, AutoTokenizer model = AutoModelForSequenceClassification.from_pretrained("dee4hf/autotrain-deephate2-1093539673", use_auth_token=True) tokenizer = AutoTokenizer.from_pretrained("dee4hf/autotrain-deephate2-1093539673", use_auth_token=True) inputs = tokenizer("I love AutoTrain", return_tensors="pt") outputs = model(**inputs) ```
1,387
ltrctelugu/bigram
null
hello
6
Dror/finetuning-sentiment-model-3000-samples
null
--- license: apache-2.0 tags: - generated_from_trainer datasets: - imdb metrics: - accuracy - f1 model-index: - name: finetuning-sentiment-model-3000-samples results: - task: name: Text Classification type: text-classification dataset: name: imdb type: imdb args: plain_text metrics: - name: Accuracy type: accuracy value: 0.87 - name: F1 type: f1 value: 0.8721311475409836 --- <!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. --> # finetuning-sentiment-model-3000-samples This model is a fine-tuned version of [distilbert-base-uncased](https://huggingface.co/distilbert-base-uncased) on the imdb dataset. It achieves the following results on the evaluation set: - Loss: 0.2979 - Accuracy: 0.87 - F1: 0.8721 ## 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: 2e-05 - train_batch_size: 16 - eval_batch_size: 16 - seed: 42 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 - lr_scheduler_type: linear - num_epochs: 2 ### Training results ### Framework versions - Transformers 4.20.1 - Pytorch 1.11.0+cu113 - Datasets 2.3.2 - Tokenizers 0.12.1
1,505
juliensimon/distilbert-imdb-mlflow
null
--- license: apache-2.0 tags: - generated_from_trainer datasets: - imdb model-index: - name: distilbert-imdb-mlflow results: [] --- <!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. --> # distilbert-imdb-mlflow This model is a fine-tuned version of [distilbert-base-cased](https://huggingface.co/distilbert-base-cased) on the imdb dataset. MLflow logs are included. To visualize them, just clone the repo and run : ``` mlflow ui ``` ## 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: 5e-05 - train_batch_size: 8 - eval_batch_size: 8 - seed: 42 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 - lr_scheduler_type: linear - num_epochs: 1 ### Training results ### Framework versions - Transformers 4.20.1 - Pytorch 1.10.0 - Datasets 2.3.2 - Tokenizers 0.12.1
1,150
rajpurkarlab/biobert-finetuned-change-classification
null
Entry not found
15
leokai/distilbert-base-uncased-finetuned-wikiandmark_epoch20
null
--- license: apache-2.0 tags: - generated_from_trainer metrics: - accuracy model-index: - name: distilbert-base-uncased-finetuned-wikiandmark_epoch20 results: [] --- <!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. --> # distilbert-base-uncased-finetuned-wikiandmark_epoch20 This model is a fine-tuned version of [distilbert-base-uncased](https://huggingface.co/distilbert-base-uncased) on an unknown dataset. It achieves the following results on the evaluation set: - Loss: 0.0561 - Accuracy: 0.9944 ## 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: 2e-05 - train_batch_size: 32 - eval_batch_size: 32 - seed: 42 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 - lr_scheduler_type: linear - num_epochs: 20 ### Training results | Training Loss | Epoch | Step | Validation Loss | Accuracy | |:-------------:|:-----:|:-----:|:---------------:|:--------:| | 0.0224 | 1.0 | 1859 | 0.0277 | 0.9919 | | 0.0103 | 2.0 | 3718 | 0.0298 | 0.9925 | | 0.0047 | 3.0 | 5577 | 0.0429 | 0.9924 | | 0.0038 | 4.0 | 7436 | 0.0569 | 0.9922 | | 0.0019 | 5.0 | 9295 | 0.0554 | 0.9936 | | 0.0028 | 6.0 | 11154 | 0.0575 | 0.9928 | | 0.002 | 7.0 | 13013 | 0.0544 | 0.9926 | | 0.0017 | 8.0 | 14872 | 0.0553 | 0.9935 | | 0.001 | 9.0 | 16731 | 0.0498 | 0.9924 | | 0.0001 | 10.0 | 18590 | 0.0398 | 0.9934 | | 0.0 | 11.0 | 20449 | 0.0617 | 0.9935 | | 0.0002 | 12.0 | 22308 | 0.0561 | 0.9944 | | 0.0002 | 13.0 | 24167 | 0.0755 | 0.9934 | | 0.0 | 14.0 | 26026 | 0.0592 | 0.9941 | | 0.0 | 15.0 | 27885 | 0.0572 | 0.9939 | | 0.0 | 16.0 | 29744 | 0.0563 | 0.9941 | | 0.0 | 17.0 | 31603 | 0.0587 | 0.9936 | | 0.0005 | 18.0 | 33462 | 0.0673 | 0.9937 | | 0.0 | 19.0 | 35321 | 0.0651 | 0.9933 | | 0.0 | 20.0 | 37180 | 0.0683 | 0.9936 | ### Framework versions - Transformers 4.20.1 - Pytorch 1.12.0+cu113 - Datasets 2.3.2 - Tokenizers 0.12.1
2,613
James-kc-min/L_Roberta3
null
--- license: apache-2.0 tags: - generated_from_trainer metrics: - accuracy - f1 - precision - recall model-index: - name: L_Roberta3 results: [] --- <!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. --> # L_Roberta3 This model is a fine-tuned version of [distilroberta-base](https://huggingface.co/distilroberta-base) on the None dataset. It achieves the following results on the evaluation set: - Loss: 0.2095 - Accuracy: 0.9555 - F1: 0.9555 - Precision: 0.9555 - Recall: 0.9555 - C Report: precision recall f1-score support 0 0.97 0.95 0.96 876 1 0.94 0.97 0.95 696 accuracy 0.96 1572 macro avg 0.95 0.96 0.96 1572 weighted avg 0.96 0.96 0.96 1572 - C Matrix: None ## 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: 5e-05 - train_batch_size: 32 - eval_batch_size: 32 - seed: 42 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 - lr_scheduler_type: linear - num_epochs: 5 - mixed_precision_training: Native AMP ### Training results | Training Loss | Epoch | Step | Validation Loss | Accuracy | F1 | Precision | Recall | C Report | C Matrix | |:-------------:|:-----:|:----:|:---------------:|:--------:|:------:|:---------:|:------:|:--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------:|:--------:| | 0.2674 | 1.0 | 329 | 0.2436 | 0.9389 | 0.9389 | 0.9389 | 0.9389 | precision recall f1-score support 0 0.94 0.95 0.95 876 1 0.94 0.92 0.93 696 accuracy 0.94 1572 macro avg 0.94 0.94 0.94 1572 weighted avg 0.94 0.94 0.94 1572 | None | | 0.1377 | 2.0 | 658 | 0.1506 | 0.9408 | 0.9408 | 0.9408 | 0.9408 | precision recall f1-score support 0 0.97 0.92 0.95 876 1 0.91 0.96 0.94 696 accuracy 0.94 1572 macro avg 0.94 0.94 0.94 1572 weighted avg 0.94 0.94 0.94 1572 | None | | 0.0898 | 3.0 | 987 | 0.1491 | 0.9548 | 0.9548 | 0.9548 | 0.9548 | precision recall f1-score support 0 0.96 0.96 0.96 876 1 0.95 0.95 0.95 696 accuracy 0.95 1572 macro avg 0.95 0.95 0.95 1572 weighted avg 0.95 0.95 0.95 1572 | None | | 0.0543 | 4.0 | 1316 | 0.1831 | 0.9561 | 0.9561 | 0.9561 | 0.9561 | precision recall f1-score support 0 0.97 0.95 0.96 876 1 0.94 0.96 0.95 696 accuracy 0.96 1572 macro avg 0.95 0.96 0.96 1572 weighted avg 0.96 0.96 0.96 1572 | None | | 0.0394 | 5.0 | 1645 | 0.2095 | 0.9555 | 0.9555 | 0.9555 | 0.9555 | precision recall f1-score support 0 0.97 0.95 0.96 876 1 0.94 0.97 0.95 696 accuracy 0.96 1572 macro avg 0.95 0.96 0.96 1572 weighted avg 0.96 0.96 0.96 1572 | None | ### Framework versions - Transformers 4.18.0 - Pytorch 1.10.2+cu102 - Datasets 2.2.2 - Tokenizers 0.12.1
4,623
anneke/finetuning-distilbert-base-uncased-finetuned-sst-2-english-5000-samples
[ "NEGATIVE", "POSITIVE" ]
--- license: apache-2.0 tags: - generated_from_trainer metrics: - accuracy - f1 model-index: - name: finetuning-distilbert-base-uncased-finetuned-sst-2-english-5000-samples results: [] --- <!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. --> # finetuning-distilbert-base-uncased-finetuned-sst-2-english-5000-samples This model is a fine-tuned version of [distilbert-base-uncased-finetuned-sst-2-english](https://huggingface.co/distilbert-base-uncased-finetuned-sst-2-english) on an unknown dataset. It achieves the following results on the evaluation set: - Loss: 0.1289 - Accuracy: 0.977 - F1: 0.9878 ## 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: 2e-05 - train_batch_size: 16 - eval_batch_size: 16 - seed: 42 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 - lr_scheduler_type: linear - num_epochs: 5 ### Training results ### Framework versions - Transformers 4.20.1 - Pytorch 1.12.0+cu113 - Datasets 2.3.2 - Tokenizers 0.12.1
1,327
johnheo1128/distilbert-base-uncased-finetuned-cola
null
--- license: apache-2.0 tags: - generated_from_trainer datasets: - glue metrics: - matthews_correlation model-index: - name: distilbert-base-uncased-finetuned-cola results: - task: name: Text Classification type: text-classification dataset: name: glue type: glue args: cola metrics: - name: Matthews Correlation type: matthews_correlation value: 0.5477951635989807 --- <!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. --> # distilbert-base-uncased-finetuned-cola This model is a fine-tuned version of [distilbert-base-uncased](https://huggingface.co/distilbert-base-uncased) on the glue dataset. It achieves the following results on the evaluation set: - Loss: 0.8081 - Matthews Correlation: 0.5478 ## 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: 2e-05 - train_batch_size: 16 - eval_batch_size: 16 - seed: 42 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 - lr_scheduler_type: linear - num_epochs: 5 ### Training results | Training Loss | Epoch | Step | Validation Loss | Matthews Correlation | |:-------------:|:-----:|:----:|:---------------:|:--------------------:| | 0.5222 | 1.0 | 535 | 0.5270 | 0.4182 | | 0.3451 | 2.0 | 1070 | 0.5017 | 0.4810 | | 0.2309 | 3.0 | 1605 | 0.5983 | 0.5314 | | 0.179 | 4.0 | 2140 | 0.7488 | 0.5291 | | 0.1328 | 5.0 | 2675 | 0.8081 | 0.5478 | ### Framework versions - Transformers 4.20.1 - Pytorch 1.12.0+cu113 - Datasets 2.3.2 - Tokenizers 0.12.1
1,999
CShorten/ArXiv-Cross-Encoder-Title-Abstracts
null
Entry not found
15
tattle-admin/july22-xlmtwtroberta-da-multi
null
Entry not found
15
SIMAS-UN/blaming_locals
null
Entry not found
15
Yuetian/bert-base-uncased-finetuned-plutchik-emotion
[ "anger", "anticipation", "disgust", "fear", "joy", "sadness", "surprise", "trust" ]
--- license: mit ---
21
09panesara/distilbert-base-uncased-finetuned-cola
null
--- license: apache-2.0 tags: - generated_from_trainer datasets: - glue metrics: - matthews_correlation model-index: - name: distilbert-base-uncased-finetuned-cola results: - task: name: Text Classification type: text-classification dataset: name: glue type: glue args: cola metrics: - name: Matthews Correlation type: matthews_correlation value: 0.5406394412669151 --- <!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. --> # distilbert-base-uncased-finetuned-cola This model is a fine-tuned version of [distilbert-base-uncased](https://huggingface.co/distilbert-base-uncased) on the glue dataset. It achieves the following results on the evaluation set: - Loss: 0.7580 - Matthews Correlation: 0.5406 ## 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: 2e-05 - train_batch_size: 16 - eval_batch_size: 16 - seed: 42 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 - lr_scheduler_type: linear - num_epochs: 5 ### Training results | Training Loss | Epoch | Step | Validation Loss | Matthews Correlation | |:-------------:|:-----:|:----:|:---------------:|:--------------------:| | 0.5307 | 1.0 | 535 | 0.5094 | 0.4152 | | 0.3545 | 2.0 | 1070 | 0.5230 | 0.4940 | | 0.2371 | 3.0 | 1605 | 0.6412 | 0.5087 | | 0.1777 | 4.0 | 2140 | 0.7580 | 0.5406 | | 0.1288 | 5.0 | 2675 | 0.8494 | 0.5396 | ### Framework versions - Transformers 4.14.1 - Pytorch 1.10.0+cu111 - Datasets 1.16.1 - Tokenizers 0.10.3
2,000
ASCCCCCCCC/distilbert-base-uncased-finetuned-clinc
[ "accept_reservations", "account_blocked", "alarm", "application_status", "apr", "are_you_a_bot", "balance", "bill_balance", "bill_due", "book_flight", "book_hotel", "calculator", "calendar", "calendar_update", "calories", "cancel", "cancel_reservation", "car_rental", "card_declin...
--- license: apache-2.0 tags: - generated_from_trainer model_index: - name: distilbert-base-uncased-finetuned-clinc results: - task: name: Text Classification type: text-classification --- <!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. --> # distilbert-base-uncased-finetuned-clinc This model is a fine-tuned version of [distilbert-base-uncased](https://huggingface.co/distilbert-base-uncased) on an unkown dataset. ## 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: 2e-05 - train_batch_size: 48 - eval_batch_size: 48 - seed: 42 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 - lr_scheduler_type: linear - num_epochs: 1 ### Framework versions - Transformers 4.9.0 - Pytorch 1.7.1+cpu - Datasets 1.17.0 - Tokenizers 0.10.3
1,130
Ahren09/distilbert-base-uncased-finetuned-cola
null
Entry not found
15
Alireza1044/albert-base-v2-qqp
null
--- language: - en license: apache-2.0 tags: - generated_from_trainer datasets: - glue metrics: - accuracy - f1 model_index: - name: qqp results: - task: name: Text Classification type: text-classification dataset: name: GLUE QQP type: glue args: qqp metric: name: F1 type: f1 value: 0.8722569490623753 --- <!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. --> # qqp This model is a fine-tuned version of [albert-base-v2](https://huggingface.co/albert-base-v2) on the GLUE QQP dataset. It achieves the following results on the evaluation set: - Loss: 0.3695 - Accuracy: 0.9050 - F1: 0.8723 - Combined Score: 0.8886 ## 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: 5e-05 - train_batch_size: 64 - eval_batch_size: 8 - seed: 42 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 - lr_scheduler_type: linear - num_epochs: 4.0 ### Training results ### Framework versions - Transformers 4.9.1 - Pytorch 1.9.0+cu102 - Datasets 1.10.2 - Tokenizers 0.10.3
1,397
Alireza1044/bert_classification_lm
null
A simple model trained on dialogues of characters in NBC series, `The Office`. The model can do a binary classification between `Michael Scott` and `Dwight Shrute`'s dialogues. <style type="text/css"> .tg {border-collapse:collapse;border-spacing:0;} .tg td{border-color:black;border-style:solid;border-width:1px;font-family:Arial, sans-serif;font-size:14px; overflow:hidden;padding:10px 5px;word-break:normal;} .tg th{border-color:black;border-style:solid;border-width:1px;font-family:Arial, sans-serif;font-size:14px; font-weight:normal;overflow:hidden;padding:10px 5px;word-break:normal;} .tg .tg-c3ow{border-color:inherit;text-align:center;vertical-align:top} </style> <table class="tg"> <thead> <tr> <th class="tg-c3ow" colspan="2">Label Definitions</th> </tr> </thead> <tbody> <tr> <td class="tg-c3ow">Label 0</td> <td class="tg-c3ow">Michael</td> </tr> <tr> <td class="tg-c3ow">Label 1</td> <td class="tg-c3ow">Dwight</td> </tr> </tbody> </table>
990
Amalq/distilbert-base-uncased-finetuned-cola
null
--- license: apache-2.0 tags: - generated_from_trainer datasets: - glue metrics: - matthews_correlation model-index: - name: distilbert-base-uncased-finetuned-cola results: - task: name: Text Classification type: text-classification dataset: name: glue type: glue args: cola metrics: - name: Matthews Correlation type: matthews_correlation value: 0.5335074704896392 --- <!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. --> # distilbert-base-uncased-finetuned-cola This model is a fine-tuned version of [distilbert-base-uncased](https://huggingface.co/distilbert-base-uncased) on the glue dataset. It achieves the following results on the evaluation set: - Loss: 0.7570 - Matthews Correlation: 0.5335 ## 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: 2e-05 - train_batch_size: 16 - eval_batch_size: 16 - seed: 42 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 - lr_scheduler_type: linear - num_epochs: 5 ### Training results | Training Loss | Epoch | Step | Validation Loss | Matthews Correlation | |:-------------:|:-----:|:----:|:---------------:|:--------------------:| | 0.5315 | 1.0 | 535 | 0.5214 | 0.4009 | | 0.354 | 2.0 | 1070 | 0.5275 | 0.4857 | | 0.2396 | 3.0 | 1605 | 0.6610 | 0.4901 | | 0.1825 | 4.0 | 2140 | 0.7570 | 0.5335 | | 0.1271 | 5.0 | 2675 | 0.8923 | 0.5074 | ### Framework versions - Transformers 4.16.2 - Pytorch 1.10.0+cu111 - Datasets 1.18.3 - Tokenizers 0.11.0
2,000
Anamika/autonlp-fa-473312409
[ "Claim", "Concluding Statement", "Counterclaim", "Evidence", "Lead", "Position", "Rebuttal" ]
--- tags: autonlp language: en widget: - text: "I love AutoNLP 🤗" datasets: - Anamika/autonlp-data-fa co2_eq_emissions: 25.128735714898614 --- # Model Trained Using AutoNLP - Problem type: Multi-class Classification - Model ID: 473312409 - CO2 Emissions (in grams): 25.128735714898614 ## Validation Metrics - Loss: 0.6010786890983582 - Accuracy: 0.7990650945370823 - Macro F1: 0.7429662929144928 - Micro F1: 0.7990650945370823 - Weighted F1: 0.7977660363770382 - Macro Precision: 0.7744390888231261 - Micro Precision: 0.7990650945370823 - Weighted Precision: 0.800444194278352 - Macro Recall: 0.7198278524814119 - Micro Recall: 0.7990650945370823 - Weighted Recall: 0.7990650945370823 ## Usage You can use cURL to access this model: ``` $ curl -X POST -H "Authorization: Bearer YOUR_API_KEY" -H "Content-Type: application/json" -d '{"inputs": "I love AutoNLP"}' https://api-inference.huggingface.co/models/Anamika/autonlp-fa-473312409 ``` Or Python API: ``` from transformers import AutoModelForSequenceClassification, AutoTokenizer model = AutoModelForSequenceClassification.from_pretrained("Anamika/autonlp-fa-473312409", use_auth_token=True) tokenizer = AutoTokenizer.from_pretrained("Anamika/autonlp-fa-473312409", use_auth_token=True) inputs = tokenizer("I love AutoNLP", return_tensors="pt") outputs = model(**inputs) ```
1,341
AnonARR/qqp-bert
[ "duplicate", "not_duplicate" ]
Entry not found
15
AnonymousSub/cline-s10-AR
null
Entry not found
15
AnonymousSub/cline_wikiqa
null
Entry not found
15
AnonymousSub/consert-s10-SR
null
Entry not found
15
AnonymousSub/declutr-emanuals-s10-AR
null
Entry not found
15
AnonymousSub/declutr-emanuals-s10-SR
null
Entry not found
15
AnonymousSub/declutr-model_wikiqa
null
Entry not found
15
AnonymousSub/declutr-s10-AR
null
Entry not found
15
AnonymousSub/declutr-s10-SR
null
Entry not found
15
AnonymousSub/dummy_1
null
Entry not found
15
AnonymousSub/rule_based_bert_quadruplet_epochs_1_shard_1_wikiqa
null
Entry not found
15
AnonymousSub/rule_based_hier_quadruplet_epochs_1_shard_1_wikiqa
null
Entry not found
15
AnonymousSub/rule_based_only_classfn_epochs_1_shard_1_wikiqa
null
Entry not found
15
AnonymousSub/rule_based_only_classfn_twostage_epochs_1_shard_1_wikiqa
null
Entry not found
15
AnonymousSub/rule_based_roberta_twostage_quadruplet_epochs_1_shard_1_wikiqa
null
Entry not found
15
AnonymousSub/rule_based_roberta_twostagetriplet_hier_epochs_1_shard_1_wikiqa
null
Entry not found
15
AnonymousSub/rule_based_twostagequadruplet_hier_epochs_1_shard_1_wikiqa
null
Entry not found
15
AnonymousSub/rule_based_twostagetriplet_epochs_1_shard_1_wikiqa
null
Entry not found
15
Ateeb/FullEmotionDetector
[ "LABEL_0", "LABEL_1", "LABEL_10", "LABEL_11", "LABEL_12", "LABEL_13", "LABEL_14", "LABEL_15", "LABEL_16", "LABEL_17", "LABEL_18", "LABEL_19", "LABEL_2", "LABEL_20", "LABEL_21", "LABEL_22", "LABEL_23", "LABEL_24", "LABEL_25", "LABEL_26", "LABEL_3", "LABEL_4", "LABEL_5", ...
Entry not found
15
CenIA/albert-base-spanish-finetuned-mldoc
[ "LABEL_0", "LABEL_1", "LABEL_2", "LABEL_3" ]
Entry not found
15
CenIA/albert-large-spanish-finetuned-xnli
[ "LABEL_0", "LABEL_1", "LABEL_2" ]
Entry not found
15
CenIA/albert-xlarge-spanish-finetuned-mldoc
[ "LABEL_0", "LABEL_1", "LABEL_2", "LABEL_3" ]
Entry not found
15
CenIA/albert-xlarge-spanish-finetuned-pawsx
null
Entry not found
15
CenIA/albert-xlarge-spanish-finetuned-xnli
[ "LABEL_0", "LABEL_1", "LABEL_2" ]
Entry not found
15
CenIA/albert-xxlarge-spanish-finetuned-mldoc
[ "LABEL_0", "LABEL_1", "LABEL_2", "LABEL_3" ]
Entry not found
15
CenIA/bert-base-spanish-wwm-cased-finetuned-pawsx
null
Entry not found
15
CenIA/distillbert-base-spanish-uncased-finetuned-pawsx
null
Entry not found
15
Cheatham/xlm-roberta-base-finetuned
[ "LABEL_0", "LABEL_1", "LABEL_2" ]
Entry not found
15
Cheatham/xlm-roberta-large-finetuned-d1
[ "LABEL_0", "LABEL_1", "LABEL_2" ]
Entry not found
15
Cheatham/xlm-roberta-large-finetuned-d12
[ "LABEL_0", "LABEL_1", "LABEL_2" ]
Entry not found
15
Cheatham/xlm-roberta-large-finetuned
[ "LABEL_0", "LABEL_1", "LABEL_2" ]
Entry not found
15
Cheatham/xlm-roberta-large-finetuned3
[ "LABEL_0", "LABEL_1", "LABEL_2" ]
Entry not found
15
CleveGreen/FieldClassifier_v2_gpt
[ "LABEL_0", "LABEL_1", "LABEL_10", "LABEL_11", "LABEL_12", "LABEL_13", "LABEL_14", "LABEL_15", "LABEL_16", "LABEL_17", "LABEL_18", "LABEL_19", "LABEL_2", "LABEL_20", "LABEL_21", "LABEL_22", "LABEL_23", "LABEL_24", "LABEL_25", "LABEL_3", "LABEL_4", "LABEL_5", "LABEL_6", "...
Entry not found
15
CodeNinja1126/test-model
[ "LABEL_0", "LABEL_1", "LABEL_10", "LABEL_11", "LABEL_12", "LABEL_13", "LABEL_14", "LABEL_15", "LABEL_16", "LABEL_17", "LABEL_18", "LABEL_19", "LABEL_2", "LABEL_20", "LABEL_21", "LABEL_22", "LABEL_23", "LABEL_24", "LABEL_25", "LABEL_26", "LABEL_27", "LABEL_28", "LABEL_29",...
Entry not found
15
DSI/TweetBasedSA
[ "LABEL_0", "LABEL_1", "LABEL_2" ]
Entry not found
15
EhsanAghazadeh/bert-based-uncased-sst2-e1
[ "negative", "positive" ]
Entry not found
15
EhsanAghazadeh/bert-based-uncased-sst2-e6
[ "negative", "positive" ]
Entry not found
15
EhsanAghazadeh/electra-base-avg-2e-5-lcc
null
Entry not found
15
EhsanAghazadeh/electra-large-lcc-2e-5-42
null
Entry not found
15
Eugenia/roberta-base-bne-finetuned-amazon_reviews_multi
null
Entry not found
15
HackMIT/double-agent
null
Entry not found
15
Hate-speech-CNERG/deoffxlmr-mono-tamil
[ "Not_offensive", "Not_in_intended_language", "Off_target_other", "Off_target_group", "Profanity", "Off_target_ind" ]
--- language: ta license: apache-2.0 --- This model is used to detect **Offensive Content** in **Tamil Code-Mixed language**. The mono in the name refers to the monolingual setting, where the model is trained using only Tamil(pure and code-mixed) data. The weights are initialized from pretrained XLM-Roberta-Base and pretrained using Masked Language Modelling on the target dataset before fine-tuning using Cross-Entropy Loss. This model is the best of multiple trained for **EACL 2021 Shared Task on Offensive Language Identification in Dravidian Languages**. Genetic-Algorithm based ensembled test predictions got the highest weighted F1 score at the leaderboard (Weighted F1 score on hold out test set: This model - 0.76, Ensemble - 0.78) ### For more details about our paper Debjoy Saha, Naman Paharia, Debajit Chakraborty, Punyajoy Saha, Animesh Mukherjee. "[Hate-Alert@DravidianLangTech-EACL2021: Ensembling strategies for Transformer-based Offensive language Detection](https://www.aclweb.org/anthology/2021.dravidianlangtech-1.38/)". ***Please cite our paper in any published work that uses any of these resources.*** ~~~ @inproceedings{saha-etal-2021-hate, title = "Hate-Alert@{D}ravidian{L}ang{T}ech-{EACL}2021: Ensembling strategies for Transformer-based Offensive language Detection", author = "Saha, Debjoy and Paharia, Naman and Chakraborty, Debajit and Saha, Punyajoy and Mukherjee, Animesh", booktitle = "Proceedings of the First Workshop on Speech and Language Technologies for Dravidian Languages", month = apr, year = "2021", address = "Kyiv", publisher = "Association for Computational Linguistics", url = "https://www.aclweb.org/anthology/2021.dravidianlangtech-1.38", pages = "270--276", abstract = "Social media often acts as breeding grounds for different forms of offensive content. For low resource languages like Tamil, the situation is more complex due to the poor performance of multilingual or language-specific models and lack of proper benchmark datasets. Based on this shared task {``}Offensive Language Identification in Dravidian Languages{''} at EACL 2021; we present an exhaustive exploration of different transformer models, We also provide a genetic algorithm technique for ensembling different models. Our ensembled models trained separately for each language secured the first position in Tamil, the second position in Kannada, and the first position in Malayalam sub-tasks. The models and codes are provided.", } ~~~
2,503
Hormigo/roberta-base-bne-finetuned-amazon_reviews_multi
null
--- license: cc-by-4.0 tags: - generated_from_trainer datasets: - amazon_reviews_multi metrics: - accuracy model_index: - name: roberta-base-bne-finetuned-amazon_reviews_multi results: - task: name: Text Classification type: text-classification dataset: name: amazon_reviews_multi type: amazon_reviews_multi args: es metric: name: Accuracy type: accuracy value: 0.9335 --- <!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. --> # roberta-base-bne-finetuned-amazon_reviews_multi This model is a fine-tuned version of [BSC-TeMU/roberta-base-bne](https://huggingface.co/BSC-TeMU/roberta-base-bne) on the amazon_reviews_multi dataset. It achieves the following results on the evaluation set: - Loss: 0.2275 - Accuracy: 0.9335 ## 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: 2e-05 - train_batch_size: 16 - eval_batch_size: 16 - seed: 42 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 - lr_scheduler_type: linear - num_epochs: 2 ### Training results | Training Loss | Epoch | Step | Validation Loss | Accuracy | |:-------------:|:-----:|:----:|:---------------:|:--------:| | 0.1909 | 1.0 | 1250 | 0.1717 | 0.9333 | | 0.0932 | 2.0 | 2500 | 0.2275 | 0.9335 | ### Framework versions - Transformers 4.9.2 - Pytorch 1.9.0+cu102 - Datasets 1.11.0 - Tokenizers 0.10.3
1,750
Huffon/qnli
null
Entry not found
15
Jeska/VaccinChatSentenceClassifierDutch_fromBERTje2_DAdialog
[ "chitchat_ask_bye", "chitchat_ask_hi", "chitchat_ask_hi_de", "chitchat_ask_hi_en", "chitchat_ask_hi_fr", "chitchat_ask_hoe_gaat_het", "chitchat_ask_name", "chitchat_ask_thanks", "faq_ask_aantal_gevaccineerd", "faq_ask_aantal_gevaccineerd_wereldwijd", "faq_ask_afspraak_afzeggen", "faq_ask_afspr...
--- tags: - generated_from_trainer metrics: - accuracy model-index: - name: VaccinChatSentenceClassifierDutch_fromBERTje2_DAdialog results: [] --- <!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. --> # VaccinChatSentenceClassifierDutch_fromBERTje2_DAdialog This model is a fine-tuned version of [outputDA/checkpoint-7710](https://huggingface.co/outputDA/checkpoint-7710) on an unknown dataset. It achieves the following results on the evaluation set: - Loss: 0.5025 - Accuracy: 0.9077 ## 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: 1e-05 - train_batch_size: 8 - eval_batch_size: 8 - seed: 42 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-06 - lr_scheduler_type: linear - num_epochs: 15.0 ### Training results | Training Loss | Epoch | Step | Validation Loss | Accuracy | |:-------------:|:-----:|:-----:|:---------------:|:--------:| | 3.9925 | 1.0 | 1320 | 3.0954 | 0.4223 | | 2.5041 | 2.0 | 2640 | 1.9762 | 0.6563 | | 1.8061 | 3.0 | 3960 | 1.3196 | 0.7952 | | 1.0694 | 4.0 | 5280 | 0.9304 | 0.8510 | | 0.6479 | 5.0 | 6600 | 0.6875 | 0.8821 | | 0.4408 | 6.0 | 7920 | 0.5692 | 0.8976 | | 0.2542 | 7.0 | 9240 | 0.5291 | 0.8949 | | 0.1709 | 8.0 | 10560 | 0.5038 | 0.9059 | | 0.1181 | 9.0 | 11880 | 0.4885 | 0.9049 | | 0.0878 | 10.0 | 13200 | 0.4900 | 0.9049 | | 0.0702 | 11.0 | 14520 | 0.4930 | 0.9086 | | 0.0528 | 12.0 | 15840 | 0.4987 | 0.9113 | | 0.0406 | 13.0 | 17160 | 0.5009 | 0.9113 | | 0.0321 | 14.0 | 18480 | 0.5017 | 0.9104 | | 0.0308 | 15.0 | 19800 | 0.5025 | 0.9077 | ### Framework versions - Transformers 4.13.0.dev0 - Pytorch 1.10.0 - Datasets 1.16.1 - Tokenizers 0.10.3
2,282
Katsiaryna/distilbert-base-uncased-finetuned
[ "LABEL_0" ]
--- license: apache-2.0 tags: - generated_from_trainer metrics: - accuracy model-index: - name: distilbert-base-uncased-finetuned results: [] --- <!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. --> # distilbert-base-uncased-finetuned This model is a fine-tuned version of [distilbert-base-uncased](https://huggingface.co/distilbert-base-uncased) on the None dataset. It achieves the following results on the evaluation set: - Loss: 0.8229 - Accuracy: 0.54 ## 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: 2e-05 - train_batch_size: 16 - eval_batch_size: 16 - seed: 42 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 - lr_scheduler_type: linear - num_epochs: 5 ### Training results | Training Loss | Epoch | Step | Validation Loss | Accuracy | |:-------------:|:-----:|:----:|:---------------:|:--------:| | No log | 1.0 | 7 | 0.7709 | 0.74 | | No log | 2.0 | 14 | 0.7048 | 0.72 | | No log | 3.0 | 21 | 0.8728 | 0.46 | | No log | 4.0 | 28 | 0.7849 | 0.64 | | No log | 5.0 | 35 | 0.8229 | 0.54 | ### Framework versions - Transformers 4.12.5 - Pytorch 1.10.0+cu111 - Datasets 1.16.1 - Tokenizers 0.10.3
1,617
Katsiaryna/qnli-electra-base-finetuned_9th_auc_ce
[ "LABEL_0" ]
Entry not found
15
Katsiaryna/qnli-electra-base-finetuned_9th_auc_ce_diff
[ "LABEL_0" ]
Entry not found
15
Katsiaryna/qnli-electra-base-finetuned_auc
[ "LABEL_0" ]
Entry not found
15
Katsiaryna/stsb-TinyBERT-L-4-finetuned_auc
[ "LABEL_0" ]
Entry not found
15
Katsiaryna/stsb-TinyBERT-L-4-finetuned_auc_151221-top3
[ "LABEL_0" ]
Entry not found
15
Katsiaryna/stsb-TinyBERT-L-4-finetuned_auc_161221-top3
[ "LABEL_0" ]
Entry not found
15
Katsiaryna/stsb-TinyBERT-L-4-finetuned_auc_40000-top3-BCE
[ "LABEL_0" ]
Entry not found
15
Kien/distilbert-base-uncased-finetuned-cola
null
--- license: apache-2.0 tags: - generated_from_trainer datasets: - glue metrics: - matthews_correlation model-index: - name: distilbert-base-uncased-finetuned-cola results: - task: name: Text Classification type: text-classification dataset: name: glue type: glue args: cola metrics: - name: Matthews Correlation type: matthews_correlation value: 0.5232819075279987 --- <!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. --> # distilbert-base-uncased-finetuned-cola This model is a fine-tuned version of [distilbert-base-uncased](https://huggingface.co/distilbert-base-uncased) on the glue dataset. It achieves the following results on the evaluation set: - Loss: 0.5327 - Matthews Correlation: 0.5233 ## 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: 2e-05 - train_batch_size: 16 - eval_batch_size: 16 - seed: 42 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 - lr_scheduler_type: linear - num_epochs: 5 ### Training results | Training Loss | Epoch | Step | Validation Loss | Matthews Correlation | |:-------------:|:-----:|:----:|:---------------:|:--------------------:| | 0.5314 | 1.0 | 535 | 0.4955 | 0.4270 | | 0.3545 | 2.0 | 1070 | 0.5327 | 0.5233 | | 0.2418 | 3.0 | 1605 | 0.6180 | 0.5132 | | 0.1722 | 4.0 | 2140 | 0.7344 | 0.5158 | | 0.1243 | 5.0 | 2675 | 0.8581 | 0.5196 | ### Framework versions - Transformers 4.15.0 - Pytorch 1.10.0+cu111 - Datasets 1.17.0 - Tokenizers 0.10.3
2,000
Kumicho/distilbert-base-uncased-finetuned-cola
null
--- license: apache-2.0 tags: - generated_from_trainer datasets: - glue metrics: - matthews_correlation model-index: - name: distilbert-base-uncased-finetuned-cola results: - task: name: Text Classification type: text-classification dataset: name: glue type: glue args: cola metrics: - name: Matthews Correlation type: matthews_correlation value: 0.5258663312307151 --- <!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. --> # distilbert-base-uncased-finetuned-cola This model is a fine-tuned version of [distilbert-base-uncased](https://huggingface.co/distilbert-base-uncased) on the glue dataset. It achieves the following results on the evaluation set: - Loss: 0.7758 - Matthews Correlation: 0.5259 ## 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: 2e-05 - train_batch_size: 16 - eval_batch_size: 16 - seed: 42 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 - lr_scheduler_type: linear - num_epochs: 1 ### Training results | Training Loss | Epoch | Step | Validation Loss | Matthews Correlation | |:-------------:|:-----:|:----:|:---------------:|:--------------------:| | 0.1926 | 1.0 | 535 | 0.7758 | 0.5259 | ### Framework versions - Transformers 4.16.2 - Pytorch 1.10.0+cu111 - Datasets 1.18.3 - Tokenizers 0.11.0
1,704
LilaBoualili/bert-pre-doc
null
Entry not found
15
LilaBoualili/electra-pre-doc
null
Entry not found
15
LilaBoualili/electra-pre-pair
null
Entry not found
15
LilaBoualili/electra-sim-doc
null
Entry not found
15
LilaBoualili/electra-vanilla
null
At its core it uses an ELECTRA-Base model (google/electra-base-discriminator) fine-tuned on the MS MARCO passage classification task. It can be loaded using the TF/AutoModelForSequenceClassification classes but it follows the same classification layer defined for BERT similarly to the TFElectraRelevanceHead in the Capreolus BERT-MaxP implementation. Refer to our [github repository](https://github.com/BOUALILILila/ExactMatchMarking) for a usage example for ad hoc ranking.
476
Lumos/imdb3_hga
null
Entry not found
15
Lumos/yahoo1
[ "LABEL_0", "LABEL_1", "LABEL_2", "LABEL_3", "LABEL_4", "LABEL_5", "LABEL_6", "LABEL_7", "LABEL_8", "LABEL_9" ]
Entry not found
15
M-FAC/bert-mini-finetuned-qqp
null
# BERT-mini model finetuned with M-FAC This model is finetuned on QQP dataset with state-of-the-art second-order optimizer M-FAC. Check NeurIPS 2021 paper for more details on M-FAC: [https://arxiv.org/pdf/2107.03356.pdf](https://arxiv.org/pdf/2107.03356.pdf). ## Finetuning setup For fair comparison against default Adam baseline, we finetune the model in the same framework as described here [https://github.com/huggingface/transformers/tree/master/examples/pytorch/text-classification](https://github.com/huggingface/transformers/tree/master/examples/pytorch/text-classification) and just swap Adam optimizer with M-FAC. Hyperparameters used by M-FAC optimizer: ```bash learning rate = 1e-4 number of gradients = 1024 dampening = 1e-6 ``` ## Results We share the best model out of 5 runs with the following score on QQP validation set: ```bash f1 = 82.98 accuracy = 87.03 ``` Mean and standard deviation for 5 runs on QQP validation set: | | F1 | Accuracy | |:----:|:-----------:|:----------:| | Adam | 82.43 ± 0.10 | 86.45 ± 0.12 | | M-FAC | 82.67 ± 0.23 | 86.75 ± 0.20 | Results can be reproduced by adding M-FAC optimizer code in [https://github.com/huggingface/transformers/blob/master/examples/pytorch/text-classification/run_glue.py](https://github.com/huggingface/transformers/blob/master/examples/pytorch/text-classification/run_glue.py) and running the following bash script: ```bash CUDA_VISIBLE_DEVICES=0 python run_glue.py \ --seed 10723 \ --model_name_or_path prajjwal1/bert-mini \ --task_name qqp \ --do_train \ --do_eval \ --max_seq_length 128 \ --per_device_train_batch_size 32 \ --learning_rate 1e-4 \ --num_train_epochs 5 \ --output_dir out_dir/ \ --optim MFAC \ --optim_args '{"lr": 1e-4, "num_grads": 1024, "damp": 1e-6}' ``` We believe these results could be improved with modest tuning of hyperparameters: `per_device_train_batch_size`, `learning_rate`, `num_train_epochs`, `num_grads` and `damp`. For the sake of fair comparison and a robust default setup we use the same hyperparameters across all models (`bert-tiny`, `bert-mini`) and all datasets (SQuAD version 2 and GLUE). Our code for M-FAC can be found here: [https://github.com/IST-DASLab/M-FAC](https://github.com/IST-DASLab/M-FAC). A step-by-step tutorial on how to integrate and use M-FAC with any repository can be found here: [https://github.com/IST-DASLab/M-FAC/tree/master/tutorials](https://github.com/IST-DASLab/M-FAC/tree/master/tutorials). ## BibTeX entry and citation info ```bibtex @article{frantar2021m, title={M-FAC: Efficient Matrix-Free Approximations of Second-Order Information}, author={Frantar, Elias and Kurtic, Eldar and Alistarh, Dan}, journal={Advances in Neural Information Processing Systems}, volume={35}, year={2021} } ```
2,785
M-FAC/bert-tiny-finetuned-qnli
null
# BERT-tiny model finetuned with M-FAC This model is finetuned on QNLI dataset with state-of-the-art second-order optimizer M-FAC. Check NeurIPS 2021 paper for more details on M-FAC: [https://arxiv.org/pdf/2107.03356.pdf](https://arxiv.org/pdf/2107.03356.pdf). ## Finetuning setup For fair comparison against default Adam baseline, we finetune the model in the same framework as described here [https://github.com/huggingface/transformers/tree/master/examples/pytorch/text-classification](https://github.com/huggingface/transformers/tree/master/examples/pytorch/text-classification) and just swap Adam optimizer with M-FAC. Hyperparameters used by M-FAC optimizer: ```bash learning rate = 1e-4 number of gradients = 1024 dampening = 1e-6 ``` ## Results We share the best model out of 5 runs with the following score on QNLI validation set: ```bash accuracy = 81.54 ``` Mean and standard deviation for 5 runs on QNLI validation set: | | Accuracy | |:----:|:-----------:| | Adam | 77.85 ± 0.15 | | M-FAC | 81.17 ± 0.43 | Results can be reproduced by adding M-FAC optimizer code in [https://github.com/huggingface/transformers/blob/master/examples/pytorch/text-classification/run_glue.py](https://github.com/huggingface/transformers/blob/master/examples/pytorch/text-classification/run_glue.py) and running the following bash script: ```bash CUDA_VISIBLE_DEVICES=0 python run_glue.py \ --seed 8276 \ --model_name_or_path prajjwal1/bert-tiny \ --task_name qnli \ --do_train \ --do_eval \ --max_seq_length 128 \ --per_device_train_batch_size 32 \ --learning_rate 1e-4 \ --num_train_epochs 5 \ --output_dir out_dir/ \ --optim MFAC \ --optim_args '{"lr": 1e-4, "num_grads": 1024, "damp": 1e-6}' ``` We believe these results could be improved with modest tuning of hyperparameters: `per_device_train_batch_size`, `learning_rate`, `num_train_epochs`, `num_grads` and `damp`. For the sake of fair comparison and a robust default setup we use the same hyperparameters across all models (`bert-tiny`, `bert-mini`) and all datasets (SQuAD version 2 and GLUE). Our code for M-FAC can be found here: [https://github.com/IST-DASLab/M-FAC](https://github.com/IST-DASLab/M-FAC). A step-by-step tutorial on how to integrate and use M-FAC with any repository can be found here: [https://github.com/IST-DASLab/M-FAC/tree/master/tutorials](https://github.com/IST-DASLab/M-FAC/tree/master/tutorials). ## BibTeX entry and citation info ```bibtex @article{frantar2021m, title={M-FAC: Efficient Matrix-Free Approximations of Second-Order Information}, author={Frantar, Elias and Kurtic, Eldar and Alistarh, Dan}, journal={Advances in Neural Information Processing Systems}, volume={35}, year={2021} } ```
2,729
Maelstrom77/roblclass
[ "LABEL_0", "LABEL_1", "LABEL_2", "LABEL_3", "LABEL_4", "LABEL_5" ]
Entry not found
15
Maha/OGBV-gender-indicbert-ta-fire20_fin
null
Entry not found
15
Maha/hin-trac2
null
Entry not found
15
Maunish/kgrouping-roberta-large
[ "LABEL_0", "LABEL_1", "LABEL_2" ]
Entry not found
15
MickyMike/0-GPT2SP-duracloud
[ "LABEL_0" ]
Entry not found
15
MickyMike/0-GPT2SP-mesos
[ "LABEL_0" ]
Entry not found
15
MickyMike/0-GPT2SP-mule
[ "LABEL_0" ]
Entry not found
15
MickyMike/00-GPT2SP-appceleratorstudio-aptanastudio
[ "LABEL_0" ]
Entry not found
15
MickyMike/00-GPT2SP-appceleratorstudio-titanium
[ "LABEL_0" ]
Entry not found
15
MickyMike/00-GPT2SP-aptanastudio-titanium
[ "LABEL_0" ]
Entry not found
15