modelId
stringlengths
4
111
lastModified
stringlengths
24
24
tags
list
pipeline_tag
stringlengths
5
30
author
stringlengths
2
34
config
null
securityStatus
null
id
stringlengths
4
111
likes
int64
0
9.53k
downloads
int64
2
73.6M
library_name
stringlengths
2
84
created
timestamp[us]
card
stringlengths
101
901k
card_len
int64
101
901k
embeddings
list
krumeto/setfit-recipe-classifer
2023-04-24T15:19:38.000Z
[ "sentence-transformers", "pytorch", "bert", "setfit", "text-classification", "arxiv:2209.11055", "license:apache-2.0", "region:us" ]
text-classification
krumeto
null
null
krumeto/setfit-recipe-classifer
1
2
sentence-transformers
2023-04-08T17:03:25
--- license: apache-2.0 tags: - setfit - sentence-transformers - text-classification pipeline_tag: text-classification --- # krumeto/setfit-recipe-classifer This is a [SetFit model](https://github.com/huggingface/setfit) that can be used for classification how difficult is a given recipe. The model has been trained using an efficient few-shot learning technique that involves: 1. Fine-tuning a [Sentence Transformer](https://www.sbert.net) with contrastive learning. 2. Training a classification head with features from the fine-tuned Sentence Transformer. ## Usage To use this model for inference, first install the SetFit library: ```bash python -m pip install setfit ``` You can then run inference as follows: ```python from setfit import SetFitModel # Download from Hub and run inference model = SetFitModel.from_pretrained("krumeto/setfit-recipe-classifer") complicated_recipe = """Ingredients: 4 ounces pancetta, diced into 1/4 inch cubes 2 1/2 to 3 pounds veal shanks (4 to 6 pieces 2 to 3 inches thick) 1/2 cup diced onion 1/2 cup diced celery 1/2 cup diced carrot 3 garlic cloves , minced 1 1/2 cups canned chopped tomatoes 1 1/2 cups chicken broth 1/2 cup dry white wine 1 bay leaf 1 sprig fresh thyme salt freshly ground black pepper all-purpose flour for dredging 2 tablespoons unsalted butter 2 tablespoons extra-virgin olive oil 4 3-inch strips of lemon zest Directions: Preheat oven to 375°F. Heat the olive oil over medium heat in a large Dutch oven. Cook pancetta until browned and crisp. Remove pancetta with a slotted spoon and transfer to a paper towel-lined plate. Season veal shanks with salt and pepper and dredge in flour. Cook the veal until browned on all sides, working in batches if necessary, then transfer to a plate. Add the onion, celery, carrot, garlic, and a pinch of salt to the Dutch oven and cook until softened. Stir in the tomatoes, chicken broth, dry white wine, bay leaf, and thyme sprig. Return the veal shanks and pancetta to the Dutch oven and bring the liquid to a simmer. Cover the pot and place it in the oven to braise for 2-2 1/2 hours, until the veal is very tender. Serve with gremolata and garnish with lemon zest strips. Note: To make gremolata, finely chop 2 tablespoons fresh parsley, 1 tablespoon grated lemon zest, and 1 garlic clove. Mix together and sprinkle over the osso buco before serving.""" # Run inference preds = model([complicated_recipe]) ``` ## BibTeX entry and citation info ```bibtex @article{https://doi.org/10.48550/arxiv.2209.11055, doi = {10.48550/ARXIV.2209.11055}, url = {https://arxiv.org/abs/2209.11055}, author = {Tunstall, Lewis and Reimers, Nils and Jo, Unso Eun Seo and Bates, Luke and Korat, Daniel and Wasserblat, Moshe and Pereg, Oren}, keywords = {Computation and Language (cs.CL), FOS: Computer and information sciences, FOS: Computer and information sciences}, title = {Efficient Few-Shot Learning Without Prompts}, publisher = {arXiv}, year = {2022}, copyright = {Creative Commons Attribution 4.0 International} } ```
3,025
[ [ 0.00263214111328125, -0.058990478515625, 0.03656005859375, 0.021270751953125, 0.00576019287109375, 0.0011701583862304688, -0.00885772705078125, -0.0198211669921875, 0.0080718994140625, 0.04791259765625, -0.026153564453125, -0.04449462890625, -0.038848876953125, ...
HeroGeonil/Hypert-medical
2023-05-27T12:53:48.000Z
[ "transformers", "pytorch", "bert", "fill-mask", "generated_from_trainer", "license:apache-2.0", "autotrain_compatible", "endpoints_compatible", "region:us" ]
fill-mask
HeroGeonil
null
null
HeroGeonil/Hypert-medical
0
2
transformers
2023-04-08T18:16:11
--- license: apache-2.0 tags: - generated_from_trainer model-index: - name: Hypernymy-Aware-BERT-Medical-v4 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. --> # Hypert-medical This model is a fine-tuned version of [bert-base-uncased](https://huggingface.co/bert-base-uncased) on an unknown 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: 5e-05 - train_batch_size: 36 - eval_batch_size: 24 - seed: 42 - gradient_accumulation_steps: 6 - total_train_batch_size: 216 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 - lr_scheduler_type: cosine - lr_scheduler_warmup_steps: 500 - training_steps: 10000 ### Training results ### Framework versions - Transformers 4.21.0 - Pytorch 1.10.1+cu111 - Datasets 2.10.1 - Tokenizers 0.12.1
1,147
[ [ -0.0303497314453125, -0.050018310546875, 0.0196075439453125, 0.008544921875, -0.0357666015625, -0.036865234375, -0.01531982421875, -0.0228729248046875, 0.0206756591796875, 0.028533935546875, -0.052947998046875, -0.042236328125, -0.049224853515625, -0.0050849...
arkadyark/dqn-SpaceInvadersNoFrameskip-v4-default-params
2023-04-08T19:07:34.000Z
[ "stable-baselines3", "SpaceInvadersNoFrameskip-v4", "deep-reinforcement-learning", "reinforcement-learning", "model-index", "region:us" ]
reinforcement-learning
arkadyark
null
null
arkadyark/dqn-SpaceInvadersNoFrameskip-v4-default-params
0
2
stable-baselines3
2023-04-08T19:06:47
--- library_name: stable-baselines3 tags: - SpaceInvadersNoFrameskip-v4 - deep-reinforcement-learning - reinforcement-learning - stable-baselines3 model-index: - name: DQN results: - task: type: reinforcement-learning name: reinforcement-learning dataset: name: SpaceInvadersNoFrameskip-v4 type: SpaceInvadersNoFrameskip-v4 metrics: - type: mean_reward value: 373.50 +/- 194.15 name: mean_reward verified: false --- # **DQN** Agent playing **SpaceInvadersNoFrameskip-v4** This is a trained model of a **DQN** agent playing **SpaceInvadersNoFrameskip-v4** using the [stable-baselines3 library](https://github.com/DLR-RM/stable-baselines3) and the [RL Zoo](https://github.com/DLR-RM/rl-baselines3-zoo). The RL Zoo is a training framework for Stable Baselines3 reinforcement learning agents, with hyperparameter optimization and pre-trained agents included. ## Usage (with SB3 RL Zoo) RL Zoo: https://github.com/DLR-RM/rl-baselines3-zoo<br/> SB3: https://github.com/DLR-RM/stable-baselines3<br/> SB3 Contrib: https://github.com/Stable-Baselines-Team/stable-baselines3-contrib Install the RL Zoo (with SB3 and SB3-Contrib): ```bash pip install rl_zoo3 ``` ``` # Download model and save it into the logs/ folder python -m rl_zoo3.load_from_hub --algo dqn --env SpaceInvadersNoFrameskip-v4 -orga arkadyark -f logs/ python -m rl_zoo3.enjoy --algo dqn --env SpaceInvadersNoFrameskip-v4 -f logs/ ``` If you installed the RL Zoo3 via pip (`pip install rl_zoo3`), from anywhere you can do: ``` python -m rl_zoo3.load_from_hub --algo dqn --env SpaceInvadersNoFrameskip-v4 -orga arkadyark -f logs/ python -m rl_zoo3.enjoy --algo dqn --env SpaceInvadersNoFrameskip-v4 -f logs/ ``` ## Training (with the RL Zoo) ``` python -m rl_zoo3.train --algo dqn --env SpaceInvadersNoFrameskip-v4 -f logs/ # Upload the model and generate video (when possible) python -m rl_zoo3.push_to_hub --algo dqn --env SpaceInvadersNoFrameskip-v4 -f logs/ -orga arkadyark ``` ## Hyperparameters ```python OrderedDict([('batch_size', 32), ('buffer_size', 100000), ('env_wrapper', ['stable_baselines3.common.atari_wrappers.AtariWrapper']), ('exploration_final_eps', 0.01), ('exploration_fraction', 0.1), ('frame_stack', 4), ('gradient_steps', 1), ('learning_rate', 0.0001), ('learning_starts', 100000), ('n_timesteps', 10000000.0), ('optimize_memory_usage', False), ('policy', 'CnnPolicy'), ('target_update_interval', 1000), ('train_freq', 4), ('normalize', False)]) ```
2,695
[ [ -0.041412353515625, -0.03607177734375, 0.0218048095703125, 0.0240325927734375, -0.01031494140625, -0.0175628662109375, 0.01241302490234375, -0.0146331787109375, 0.01288604736328125, 0.0246429443359375, -0.07080078125, -0.03546142578125, -0.027099609375, -0.0...
ratish/bert-textClassification_v1.1
2023-04-08T22:39:55.000Z
[ "transformers", "tf", "distilbert", "text-classification", "generated_from_keras_callback", "license:apache-2.0", "endpoints_compatible", "region:us" ]
text-classification
ratish
null
null
ratish/bert-textClassification_v1.1
0
2
transformers
2023-04-08T21:13:32
--- license: apache-2.0 tags: - generated_from_keras_callback model-index: - name: ratish/bert-textClassification_v1.1 results: [] --- <!-- This model card has been generated automatically according to the information Keras had access to. You should probably proofread and complete it, then remove this comment. --> # ratish/bert-textClassification_v1.1 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: - Train Loss: 1.2176 - Validation Loss: 1.4740 - Train Accuracy: 0.5909 - Epoch: 4 ## 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: - optimizer: {'name': 'Adam', 'weight_decay': None, 'clipnorm': None, 'global_clipnorm': None, 'clipvalue': None, 'use_ema': False, 'ema_momentum': 0.99, 'ema_overwrite_frequency': None, 'jit_compile': False, 'is_legacy_optimizer': False, 'learning_rate': {'class_name': 'PolynomialDecay', 'config': {'initial_learning_rate': 2e-05, 'decay_steps': 95, 'end_learning_rate': 0.0, 'power': 1.0, 'cycle': False, 'name': None}}, 'beta_1': 0.9, 'beta_2': 0.999, 'epsilon': 1e-08, 'amsgrad': False} - training_precision: float32 ### Training results | Train Loss | Validation Loss | Train Accuracy | Epoch | |:----------:|:---------------:|:--------------:|:-----:| | 2.2620 | 2.1136 | 0.3636 | 0 | | 1.8161 | 1.8166 | 0.3864 | 1 | | 1.4886 | 1.6061 | 0.5909 | 2 | | 1.2862 | 1.5037 | 0.5909 | 3 | | 1.2176 | 1.4740 | 0.5909 | 4 | ### Framework versions - Transformers 4.27.4 - TensorFlow 2.12.0 - Datasets 2.11.0 - Tokenizers 0.13.3
1,964
[ [ -0.039825439453125, -0.0386962890625, 0.0242919921875, 0.00545501708984375, -0.024627685546875, -0.0153961181640625, -0.021209716796875, -0.017791748046875, 0.00677490234375, -0.0029277801513671875, -0.052947998046875, -0.049224853515625, -0.059600830078125, ...
lenayagaf/bert-buzzfeed-balanced
2023-04-09T11:22:02.000Z
[ "transformers", "pytorch", "tensorboard", "bert", "text-classification", "generated_from_trainer", "license:apache-2.0", "endpoints_compatible", "region:us" ]
text-classification
lenayagaf
null
null
lenayagaf/bert-buzzfeed-balanced
0
2
transformers
2023-04-08T21:36:35
--- license: apache-2.0 tags: - generated_from_trainer metrics: - accuracy - f1 - precision - recall model-index: - name: bert-buzzfeed-balanced 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. --> # bert-buzzfeed-balanced This model is a fine-tuned version of [bert-base-uncased](https://huggingface.co/bert-base-uncased) on the None dataset. It achieves the following results on the evaluation set: - Loss: 0.6343 - Accuracy: 0.6383 - F1: 0.6383 - Precision: 0.6818 - Recall: 0.6 ## 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 | F1 | Precision | Recall | |:-------------:|:-----:|:----:|:---------------:|:--------:|:------:|:---------:|:------:| | No log | 1.0 | 47 | 0.6265 | 0.6330 | 0.6333 | 0.6667 | 0.62 | | No log | 2.0 | 94 | 0.6343 | 0.6383 | 0.6383 | 0.6818 | 0.6 | ### Framework versions - Transformers 4.27.4 - Pytorch 2.0.0+cu118 - Datasets 2.11.0 - Tokenizers 0.13.3
1,591
[ [ -0.032989501953125, -0.04473876953125, 0.00858306884765625, 0.0241241455078125, -0.0224456787109375, -0.0281829833984375, -0.018341064453125, -0.02862548828125, 0.0198822021484375, 0.014678955078125, -0.052734375, -0.036376953125, -0.048675537109375, -0.0244...
sadia72/roberta-base-finetuned-sarcasm-news-headline-detection
2023-04-08T22:11:21.000Z
[ "transformers", "pytorch", "tensorboard", "roberta", "text-classification", "generated_from_trainer", "license:mit", "endpoints_compatible", "region:us" ]
text-classification
sadia72
null
null
sadia72/roberta-base-finetuned-sarcasm-news-headline-detection
0
2
transformers
2023-04-08T21:56:54
--- license: mit tags: - generated_from_trainer model-index: - name: roberta-base-finetuned-sarcasm-news-headline-detection 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. --> # roberta-base-finetuned-sarcasm-news-headline-detection This model is a fine-tuned version of [roberta-base](https://huggingface.co/roberta-base) on an unknown dataset. It achieves the following results on the evaluation set: - Loss: 0.0451 ## 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: 8 - seed: 42 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 - lr_scheduler_type: linear - num_epochs: 3 ### Training results | Training Loss | Epoch | Step | Validation Loss | |:-------------:|:-----:|:----:|:---------------:| | 0.2325 | 1.0 | 1789 | 0.1235 | | 0.1525 | 2.0 | 3578 | 0.0767 | | 0.0944 | 3.0 | 5367 | 0.0451 | ### Framework versions - Transformers 4.27.4 - Pytorch 2.0.0+cu118 - Datasets 2.11.0 - Tokenizers 0.13.3
1,414
[ [ -0.028594970703125, -0.0533447265625, 0.0217742919921875, 0.00696563720703125, -0.0249786376953125, -0.03302001953125, -0.01910400390625, -0.01214599609375, 0.0008502006530761719, 0.034515380859375, -0.060150146484375, -0.045440673828125, -0.05615234375, -0....
approach0/splade_all-cocomae-220
2023-04-08T23:36:05.000Z
[ "transformers", "pytorch", "tensorboard", "bert", "pretraining", "azbert", "fill-mask", "en", "license:mit", "endpoints_compatible", "region:us" ]
fill-mask
approach0
null
null
approach0/splade_all-cocomae-220
0
2
transformers
2023-04-08T23:35:58
--- language: en tags: - azbert - pretraining - fill-mask widget: - text: "$f$ $($ $x$ [MASK] $y$ $)$" example_title: "mathy" - text: "$x$ [MASK] $x$ $equal$ $2$ $x$" example_title: "mathy" - text: "Proof by [MASK] that $n$ $fact$ $gt$ $3$ $n$ for $n$ $gt$ $6$" example_title: "mathy" - text: "Proof by induction that $n$ [MASK] $gt$ $3$ $n$ for $n$ $gt$ $6$" example_title: "mathy" - text: "The goal of life is [MASK]." example_title: "philosophical" license: mit --- ## About This [repository](https://github.com/approach0/azbert) is a boilerplate to push a mask-filling model to the HuggingFace Model Hub. ### Upload to huggingface Download your tokenizer, model checkpoints, and optionally the training logs (`events.out.*`) to the `./ckpt` directory (do not include any large files except `pytorch_model.bin` and log files `events.out.*`). Optionally, test model using the MLM task: ```sh pip install pya0 # for math token preprocessing # testing local checkpoints: python test.py ./ckpt/math-tokenizer ./ckpt/2-2-0/encoder.ckpt # testing Model Hub checkpoints: python test.py approach0/coco-mae-220 approach0/coco-mae-220 ``` > **Note** > Modify the test examples in `test.txt` to play with it. > The test file is tab-separated, the first column is additional positions you want to mask for the right-side sentence (useful for masking tokens in math markups). > A zero means no additional mask positions. To upload to huggingface, use the `upload2hgf.sh` script. Before runnig this script, be sure to check: * `git-lfs` is installed * having git-remote named `hgf` reference to `https://huggingface.co/your/repo` * model contains all the files needed: `config.json` and `pytorch_model.bin` * tokenizer contains all the files needed: `added_tokens.json`, `special_tokens_map.json`, `tokenizer_config.json`, `vocab.txt` and `tokenizer.json` * no `tokenizer_file` field in `tokenizer_config.json` (sometimes it is located locally at `~/.cache`)
1,964
[ [ -0.03790283203125, -0.04705810546875, 0.0060577392578125, 0.0325927734375, -0.0110321044921875, -0.0008606910705566406, 0.00016105175018310547, -0.016815185546875, 0.03887939453125, 0.04632568359375, -0.049530029296875, -0.0406494140625, -0.05340576171875, -...
ratish/bert-textClassification_v1.4
2023-04-09T00:26:36.000Z
[ "transformers", "tf", "distilbert", "text-classification", "generated_from_keras_callback", "license:apache-2.0", "endpoints_compatible", "region:us" ]
text-classification
ratish
null
null
ratish/bert-textClassification_v1.4
0
2
transformers
2023-04-09T00:17:53
--- license: apache-2.0 tags: - generated_from_keras_callback model-index: - name: ratish/bert-textClassification_v1.4 results: [] --- <!-- This model card has been generated automatically according to the information Keras had access to. You should probably proofread and complete it, then remove this comment. --> # ratish/bert-textClassification_v1.4 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: - Train Loss: 0.3431 - Validation Loss: 0.8618 - Train Accuracy: 0.7273 - Epoch: 14 ## 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: - optimizer: {'name': 'Adam', 'weight_decay': None, 'clipnorm': None, 'global_clipnorm': None, 'clipvalue': None, 'use_ema': False, 'ema_momentum': 0.99, 'ema_overwrite_frequency': None, 'jit_compile': True, 'is_legacy_optimizer': False, 'learning_rate': {'class_name': 'PolynomialDecay', 'config': {'initial_learning_rate': 2e-05, 'decay_steps': 285, 'end_learning_rate': 0.0, 'power': 1.0, 'cycle': False, 'name': None}}, 'beta_1': 0.9, 'beta_2': 0.999, 'epsilon': 1e-08, 'amsgrad': False} - training_precision: float32 ### Training results | Train Loss | Validation Loss | Train Accuracy | Epoch | |:----------:|:---------------:|:--------------:|:-----:| | 2.2087 | 1.9909 | 0.4091 | 0 | | 1.7130 | 1.6444 | 0.5909 | 1 | | 1.3350 | 1.3844 | 0.5455 | 2 | | 1.0642 | 1.2276 | 0.6136 | 3 | | 0.8599 | 1.1036 | 0.6818 | 4 | | 0.7216 | 1.0790 | 0.6818 | 5 | | 0.6305 | 1.0403 | 0.6818 | 6 | | 0.5304 | 0.9581 | 0.7045 | 7 | | 0.4899 | 0.8977 | 0.7273 | 8 | | 0.4332 | 0.8907 | 0.7273 | 9 | | 0.4000 | 0.9072 | 0.7273 | 10 | | 0.3740 | 0.8734 | 0.7273 | 11 | | 0.3579 | 0.8726 | 0.7273 | 12 | | 0.3448 | 0.8648 | 0.7273 | 13 | | 0.3431 | 0.8618 | 0.7273 | 14 | ### Framework versions - Transformers 4.27.4 - TensorFlow 2.12.0 - Datasets 2.11.0 - Tokenizers 0.13.3
2,545
[ [ -0.045623779296875, -0.038665771484375, 0.0236663818359375, 0.0032863616943359375, -0.017120361328125, -0.01250457763671875, -0.01361846923828125, -0.01499176025390625, 0.017242431640625, 0.00432586669921875, -0.0533447265625, -0.05126953125, -0.055908203125, ...
erosendo/dqn-SpaceInvaders
2023-04-09T03:48:25.000Z
[ "stable-baselines3", "SpaceInvadersNoFrameskip-v4", "deep-reinforcement-learning", "reinforcement-learning", "model-index", "region:us" ]
reinforcement-learning
erosendo
null
null
erosendo/dqn-SpaceInvaders
0
2
stable-baselines3
2023-04-09T03:47:47
--- library_name: stable-baselines3 tags: - SpaceInvadersNoFrameskip-v4 - deep-reinforcement-learning - reinforcement-learning - stable-baselines3 model-index: - name: DQN results: - task: type: reinforcement-learning name: reinforcement-learning dataset: name: SpaceInvadersNoFrameskip-v4 type: SpaceInvadersNoFrameskip-v4 metrics: - type: mean_reward value: 493.50 +/- 106.33 name: mean_reward verified: false --- # **DQN** Agent playing **SpaceInvadersNoFrameskip-v4** This is a trained model of a **DQN** agent playing **SpaceInvadersNoFrameskip-v4** using the [stable-baselines3 library](https://github.com/DLR-RM/stable-baselines3) and the [RL Zoo](https://github.com/DLR-RM/rl-baselines3-zoo). The RL Zoo is a training framework for Stable Baselines3 reinforcement learning agents, with hyperparameter optimization and pre-trained agents included. ## Usage (with SB3 RL Zoo) RL Zoo: https://github.com/DLR-RM/rl-baselines3-zoo<br/> SB3: https://github.com/DLR-RM/stable-baselines3<br/> SB3 Contrib: https://github.com/Stable-Baselines-Team/stable-baselines3-contrib Install the RL Zoo (with SB3 and SB3-Contrib): ```bash pip install rl_zoo3 ``` ``` # Download model and save it into the logs/ folder python -m rl_zoo3.load_from_hub --algo dqn --env SpaceInvadersNoFrameskip-v4 -orga erosendo -f logs/ python -m rl_zoo3.enjoy --algo dqn --env SpaceInvadersNoFrameskip-v4 -f logs/ ``` If you installed the RL Zoo3 via pip (`pip install rl_zoo3`), from anywhere you can do: ``` python -m rl_zoo3.load_from_hub --algo dqn --env SpaceInvadersNoFrameskip-v4 -orga erosendo -f logs/ python -m rl_zoo3.enjoy --algo dqn --env SpaceInvadersNoFrameskip-v4 -f logs/ ``` ## Training (with the RL Zoo) ``` python -m rl_zoo3.train --algo dqn --env SpaceInvadersNoFrameskip-v4 -f logs/ # Upload the model and generate video (when possible) python -m rl_zoo3.push_to_hub --algo dqn --env SpaceInvadersNoFrameskip-v4 -f logs/ -orga erosendo ``` ## Hyperparameters ```python OrderedDict([('batch_size', 32), ('buffer_size', 100000), ('env_wrapper', ['stable_baselines3.common.atari_wrappers.AtariWrapper']), ('exploration_final_eps', 0.01), ('exploration_fraction', 0.1), ('frame_stack', 4), ('gradient_steps', 1), ('learning_rate', 0.0001), ('learning_starts', 100000), ('n_timesteps', 1000000.0), ('optimize_memory_usage', False), ('policy', 'CnnPolicy'), ('target_update_interval', 1000), ('train_freq', 4), ('normalize', False)]) ```
2,691
[ [ -0.04217529296875, -0.037567138671875, 0.0215606689453125, 0.0253753662109375, -0.00954437255859375, -0.0189208984375, 0.013092041015625, -0.01447296142578125, 0.01264190673828125, 0.0244598388671875, -0.069091796875, -0.034515380859375, -0.0265960693359375, ...
0x7194633/roberta-base-value-determinator
2023-04-09T07:56:33.000Z
[ "transformers", "pytorch", "roberta", "text-classification", "generated_from_trainer", "license:mit", "endpoints_compatible", "region:us" ]
text-classification
0x7194633
null
null
0x7194633/roberta-base-value-determinator
0
2
transformers
2023-04-09T07:15:32
--- license: mit tags: - generated_from_trainer metrics: - accuracy - f1 - precision - recall model-index: - name: roberta-base-value-determinator 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. --> # roberta-base-value-determinator This model is a fine-tuned version of [roberta-base](https://huggingface.co/roberta-base) on the 0x7194633/value_determinant dataset. It achieves the following results on the evaluation set: - Loss: 0.0000 - Accuracy: 1.0 - F1: 1.0 - Precision: 1.0 - Recall: 1.0 ## 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: 16 - seed: 42 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 - lr_scheduler_type: linear - lr_scheduler_warmup_steps: 500 - num_epochs: 1 ### Training results ### Framework versions - Transformers 4.27.4 - Pytorch 2.0.0+cu118 - Datasets 2.11.0 - Tokenizers 0.13.3
1,270
[ [ -0.02703857421875, -0.050567626953125, 0.0157470703125, 0.002445220947265625, -0.0303497314453125, -0.0280609130859375, -0.0168609619140625, -0.0143280029296875, 0.007648468017578125, 0.0244140625, -0.049560546875, -0.039794921875, -0.05889892578125, -0.0055...
doggylion/distilbert-base-uncased-finetuned-emotion
2023-04-09T15:54:32.000Z
[ "transformers", "pytorch", "tensorboard", "distilbert", "text-classification", "generated_from_trainer", "dataset:emotion", "license:apache-2.0", "model-index", "endpoints_compatible", "region:us" ]
text-classification
doggylion
null
null
doggylion/distilbert-base-uncased-finetuned-emotion
0
2
transformers
2023-04-09T12:09:50
--- 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 config: split split: validation args: split metrics: - name: Accuracy type: accuracy value: 0.924 - name: F1 type: f1 value: 0.9241955876397631 --- <!-- 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: 0.2180 - Accuracy: 0.924 - F1: 0.9242 ## 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 | |:-------------:|:-----:|:----:|:---------------:|:--------:|:------:| | 0.8185 | 1.0 | 250 | 0.3127 | 0.9035 | 0.9002 | | 0.2449 | 2.0 | 500 | 0.2180 | 0.924 | 0.9242 | ### Framework versions - Transformers 4.27.4 - Pytorch 2.0.0+cu118 - Datasets 2.11.0 - Tokenizers 0.13.3
1,846
[ [ -0.037811279296875, -0.041259765625, 0.01416778564453125, 0.0219268798828125, -0.025634765625, -0.0188751220703125, -0.01325225830078125, -0.00885772705078125, 0.01067352294921875, 0.00830078125, -0.056121826171875, -0.0518798828125, -0.0601806640625, -0.008...
XYang2023/distilbert-base-uncased-emotion
2023-04-10T00:34:56.000Z
[ "transformers", "pytorch", "tensorboard", "distilbert", "text-classification", "generated_from_trainer", "dataset:emotion", "license:apache-2.0", "model-index", "endpoints_compatible", "region:us" ]
text-classification
XYang2023
null
null
XYang2023/distilbert-base-uncased-emotion
0
2
transformers
2023-04-09T13:28:45
--- license: apache-2.0 tags: - generated_from_trainer datasets: - emotion metrics: - f1 - accuracy model-index: - name: distilbert-base-uncased-emotion results: - task: name: Text Classification type: text-classification dataset: name: emotion type: emotion config: split split: validation args: split metrics: - name: F1 type: f1 value: 0.9200802440853002 - name: Accuracy type: accuracy value: 0.92 --- <!-- 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-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: 0.2316 - F1: 0.9201 - Accuracy: 0.92 ## 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 | F1 | Accuracy | |:-------------:|:-----:|:----:|:---------------:|:------:|:--------:| | No log | 1.0 | 250 | 0.3294 | 0.9004 | 0.903 | | No log | 2.0 | 500 | 0.2316 | 0.9201 | 0.92 | ### Framework versions - Transformers 4.27.4 - Pytorch 2.0.0 - Datasets 2.11.0 - Tokenizers 0.12.1
1,818
[ [ -0.03460693359375, -0.039520263671875, 0.0164031982421875, 0.0259246826171875, -0.0275421142578125, -0.0167388916015625, -0.0121002197265625, -0.0088043212890625, 0.0121002197265625, 0.00997161865234375, -0.056854248046875, -0.05145263671875, -0.05963134765625, ...
andreaskoepf/pythia-6.9b-gpt4all-pretrain
2023-04-09T14:33:08.000Z
[ "transformers", "pytorch", "gpt_neox", "text-generation", "license:apache-2.0", "endpoints_compatible", "has_space", "text-generation-inference", "region:us" ]
text-generation
andreaskoepf
null
null
andreaskoepf/pythia-6.9b-gpt4all-pretrain
2
2
transformers
2023-04-09T13:35:52
--- license: apache-2.0 --- wandb: https://wandb.ai/open-assistant/supervised-finetuning/runs/kzy0gark datasets: ``` pretrain: num_train_epochs: 1 weight_decay: 0.0 use_custom_sampler: true sort_by_length: false datasets: - joke - webgpt: val_split: 0.1 - gpt4all: val_split: 0.01 - alpaca: val_split: 0.025 - code_alpaca: val_split: 0.05 - minimath - humaneval_mbpp_codegen_qa - humaneval_mbpp_testgen_qa - grade_school_math_instructions - recipes - cmu_wiki_qa - oa_wiki_qa_bart_10000row - prosocial_dialogue: fraction: 0.1 - explain_prosocial: fraction: 0.05 - oig_file: source_url: https://huggingface.co/datasets/laion/OIG/resolve/main/unified_chip2.jsonl max_count: 10000 min_length: 250 val_split: 0.1 ``` pythia: ``` pythia-6.9b-pretrain: learning_rate: 6e-6 model_name: EleutherAI/pythia-6.9b-deduped deepspeed_config: configs/zero3_config_pretrain.json weight_decay: 0.0 max_length: 2048 use_flash_attention: true warmup_steps: 20 gradient_checkpointing: false gradient_accumulation_steps: 2 per_device_train_batch_size: 5 per_device_eval_batch_size: 8 num_train_epochs: 1 save_total_limit: 2 ``` command: `deepspeed trainer_sft.py --configs defaults pretrain pythia-6.9b-pretrain --cache_dir .cache/ --output_dir .saved_models/pythia-6.9b-pre --residual_dropout 0.0 --deepspeed`
1,470
[ [ -0.057464599609375, -0.056060791015625, 0.01800537109375, 0.0170745849609375, -0.0211944580078125, -0.0182037353515625, -0.016845703125, 0.003940582275390625, 0.01247406005859375, 0.0303192138671875, -0.061370849609375, -0.036346435546875, -0.045684814453125, ...
Bahasalab/BahasaGpt-chat
2023-04-11T07:23:12.000Z
[ "transformers", "pytorch", "tensorboard", "license:cc-by-nc-3.0", "endpoints_compatible", "region:us" ]
null
Bahasalab
null
null
Bahasalab/BahasaGpt-chat
1
2
transformers
2023-04-09T13:44:42
--- license: cc-by-nc-3.0 --- # BahasaGPT-Chat ## Introduction This document provides an overview of the BahasaGPT-Chat model, which is a fine-tuned model for a specific task in the Indonesian language. The model is based on the Bloomz-7B-mt architecture and is fine-tuned using a dataset of over 120000 Chat instructions based. ## Model Details **Model Name:** BahasaGPT-Chat **Model Source:** Bloomz-7B-mt **Dataset for Fine-Tuning:** Over 120k Indonesia Instruct Dataset generated using the Alpaca method from the following sources: - [Stanford Alpaca](https://github.com/tatsu-lab/stanford_alpaca) - [Baize-Chatbot] (https://github.com/project-baize/baize-chatbot) - Translated instructions from OA ([Anh/data at main · LAION-AI/Anh](https://github.com/LAION-AI/Anh)) ## Fine-Tuning Process The BahasaGPT-1 model was fine-tuned using a dataset of over 120k Indonesian instructions, which were generated using [Baize-Chatbot] (https://github.com/project-baize/baize-chatbot) method with addition alpaca and OA Translation dataset. This combination of datasets allowed the model to be better adapted to the specific needs of Indonesian language tasks. The fine-tuning process involved adjusting the model's weights and biases based on the input dataset. This was done iteratively to optimize the model's performance for the specific task in the Indonesian language. ## Known Limitations Despite the successful fine-tuning, the BahasaGPT-1 model still has some limitations: **Hallucination:** The model sometimes generates outputs that may seem plausible but are not based on the input data. This may lead to incorrect or nonsensical responses in some cases. **Bias:** The BahasaGPT-1 model, like other AI language models, can exhibit various forms of bias due to the data it was trained on. This includes, but is not limited to, gender, racial, and cultural biases. As a result, the model may generate outputs that perpetuate stereotypes, exhibit unfair treatment, or show preference for specific groups or perspectives. Efforts have been made to mitigate these biases, but they may still be present in the model's responses. ## Conclusion The BahasaGPT-1 model is a fine-tuned language model for Indonesian language tasks, based on the Bloomz-7B-mt architecture. The model was trained on a dataset of over 120k Indonesian instructions generated using using [Baize-Chatbot] (https://github.com/project-baize/baize-chatbot) method with addition alpaca and OA Translation dataset. Despite some limitations, such as occasional hallucination, the model provides a valuable tool for working with Indonesian language tasks. ## How to Run For Gradio Demo : [Gradio Code](https://github.com/acul3/Bahasa_Chat) For Colab Using (Int8) : [Colab](https://colab.research.google.com/drive/1yvhJENcd0NKuMZNipAJVP4eP-k7-ilXj?usp=sharing)
2,844
[ [ -0.03790283203125, -0.09820556640625, 0.0011310577392578125, 0.04107666015625, -0.0191497802734375, -0.0251312255859375, -0.02178955078125, -0.034820556640625, 0.01094818115234375, 0.053955078125, -0.058135986328125, -0.0311431884765625, -0.050384521484375, ...
Mukhtadir/distilbert-base-uncased-finetuned-emotion
2023-04-10T10:48:59.000Z
[ "transformers", "pytorch", "tensorboard", "distilbert", "text-classification", "generated_from_trainer", "dataset:emotion", "license:apache-2.0", "model-index", "endpoints_compatible", "region:us" ]
text-classification
Mukhtadir
null
null
Mukhtadir/distilbert-base-uncased-finetuned-emotion
0
2
transformers
2023-04-09T14:01:21
--- 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 config: split split: validation args: split metrics: - name: Accuracy type: accuracy value: 0.9275 - name: F1 type: f1 value: 0.9276531435070997 --- <!-- 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: 0.2144 - Accuracy: 0.9275 - F1: 0.9277 ## 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 | |:-------------:|:-----:|:----:|:---------------:|:--------:|:------:| | 0.8335 | 1.0 | 250 | 0.3113 | 0.904 | 0.9007 | | 0.2492 | 2.0 | 500 | 0.2144 | 0.9275 | 0.9277 | ### Framework versions - Transformers 4.27.4 - Pytorch 2.0.0+cu118 - Datasets 2.11.0 - Tokenizers 0.13.3
1,848
[ [ -0.03802490234375, -0.041351318359375, 0.015533447265625, 0.021636962890625, -0.0262603759765625, -0.0191650390625, -0.0130615234375, -0.0086212158203125, 0.01030731201171875, 0.00864410400390625, -0.055908203125, -0.051483154296875, -0.059722900390625, -0.0...
rukshanCodeGen/dummp
2023-04-09T17:44:44.000Z
[ "transformers", "pytorch", "tensorboard", "bert", "text-classification", "generated_from_trainer", "dataset:yelp_review_full", "license:apache-2.0", "model-index", "endpoints_compatible", "region:us" ]
text-classification
rukshanCodeGen
null
null
rukshanCodeGen/dummp
0
2
transformers
2023-04-09T14:20:13
--- license: apache-2.0 tags: - generated_from_trainer datasets: - yelp_review_full metrics: - accuracy model-index: - name: dummp results: - task: name: Text Classification type: text-classification dataset: name: yelp_review_full type: yelp_review_full config: yelp_review_full split: test args: yelp_review_full metrics: - name: Accuracy type: accuracy value: 0.638 --- <!-- 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. --> # dummp This model is a fine-tuned version of [bert-base-cased](https://huggingface.co/bert-base-cased) on the yelp_review_full dataset. It achieves the following results on the evaluation set: - Loss: 1.4199 - Accuracy: 0.638 ## 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: 8 - eval_batch_size: 64 - 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 | 125 | 0.9293 | 0.607 | | No log | 2.0 | 250 | 1.0291 | 0.626 | | No log | 3.0 | 375 | 1.2118 | 0.628 | | No log | 4.0 | 500 | 1.3472 | 0.633 | | No log | 5.0 | 625 | 1.4199 | 0.638 | ### Framework versions - Transformers 4.27.4 - Pytorch 2.0.0+cu118 - Datasets 2.11.0 - Tokenizers 0.13.3
1,879
[ [ -0.032318115234375, -0.043365478515625, 0.015167236328125, 0.00855255126953125, -0.0250244140625, -0.036407470703125, -0.01322174072265625, -0.0194854736328125, 0.01148223876953125, 0.0234222412109375, -0.059661865234375, -0.046966552734375, -0.041229248046875, ...
madmancity/revmlc
2023-04-11T18:21:02.000Z
[ "transformers", "pytorch", "deberta-v2", "text-classification", "sentiment-analysis", "en", "dataset:madmancity/revmlc", "endpoints_compatible", "region:us" ]
text-classification
madmancity
null
null
madmancity/revmlc
0
2
transformers
2023-04-09T14:40:11
--- tags: - text-classification - sentiment-analysis language: - en widget: - text: "I love this product! One of my best purchases this year." datasets: - madmancity/revmlc --- ## Validation Metrics - Loss: 0.595 - Accuracy: 0.789 - Macro F1: 0.575 - Micro F1: 0.789 - Weighted F1: 0.763 - Macro Precision: 0.630 - Micro Precision: 0.789 - Weighted Precision: 0.775 - Macro Recall: 0.588 - Micro Recall: 0.789 - Weighted Recall: 0.789 ## 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 this product! One of my best purchases this year."}' https://api-inference.huggingface.co/models/madmancity/revmlc ``` Or Python API: ``` from transformers import AutoModelForSequenceClassification, AutoTokenizer model = AutoModelForSequenceClassification.from_pretrained("madmancity/revmlc", use_auth_token=True) tokenizer = AutoTokenizer.from_pretrained("madmancity/revmlc", use_auth_token=True) inputs = tokenizer("I love this product! One of my best purchases this year.", return_tensors="pt") outputs = model(**inputs) ```
1,141
[ [ -0.036865234375, -0.038330078125, 0.011871337890625, 0.0323486328125, -0.002285003662109375, -0.0032482147216796875, -0.004611968994140625, -0.00974273681640625, 0.01418304443359375, 0.0102386474609375, -0.0601806640625, -0.058380126953125, -0.04351806640625, ...
ItchyB/dqn-SpaceInvadersNoFrameskip-v4
2023-04-09T20:08:40.000Z
[ "stable-baselines3", "SpaceInvadersNoFrameskip-v4", "deep-reinforcement-learning", "reinforcement-learning", "model-index", "region:us" ]
reinforcement-learning
ItchyB
null
null
ItchyB/dqn-SpaceInvadersNoFrameskip-v4
0
2
stable-baselines3
2023-04-09T15:17:38
--- library_name: stable-baselines3 tags: - SpaceInvadersNoFrameskip-v4 - deep-reinforcement-learning - reinforcement-learning - stable-baselines3 model-index: - name: DQN results: - task: type: reinforcement-learning name: reinforcement-learning dataset: name: SpaceInvadersNoFrameskip-v4 type: SpaceInvadersNoFrameskip-v4 metrics: - type: mean_reward value: 274.50 +/- 31.50 name: mean_reward verified: false --- # **DQN** Agent playing **SpaceInvadersNoFrameskip-v4** This is a trained model of a **DQN** agent playing **SpaceInvadersNoFrameskip-v4** using the [stable-baselines3 library](https://github.com/DLR-RM/stable-baselines3) and the [RL Zoo](https://github.com/DLR-RM/rl-baselines3-zoo). The RL Zoo is a training framework for Stable Baselines3 reinforcement learning agents, with hyperparameter optimization and pre-trained agents included. ## Usage (with SB3 RL Zoo) RL Zoo: https://github.com/DLR-RM/rl-baselines3-zoo<br/> SB3: https://github.com/DLR-RM/stable-baselines3<br/> SB3 Contrib: https://github.com/Stable-Baselines-Team/stable-baselines3-contrib Install the RL Zoo (with SB3 and SB3-Contrib): ```bash pip install rl_zoo3 ``` ``` # Download model and save it into the logs/ folder python -m rl_zoo3.load_from_hub --algo dqn --env SpaceInvadersNoFrameskip-v4 -orga ItchyB -f logs/ python -m rl_zoo3.enjoy --algo dqn --env SpaceInvadersNoFrameskip-v4 -f logs/ ``` If you installed the RL Zoo3 via pip (`pip install rl_zoo3`), from anywhere you can do: ``` python -m rl_zoo3.load_from_hub --algo dqn --env SpaceInvadersNoFrameskip-v4 -orga ItchyB -f logs/ python -m rl_zoo3.enjoy --algo dqn --env SpaceInvadersNoFrameskip-v4 -f logs/ ``` ## Training (with the RL Zoo) ``` python -m rl_zoo3.train --algo dqn --env SpaceInvadersNoFrameskip-v4 -f logs/ # Upload the model and generate video (when possible) python -m rl_zoo3.push_to_hub --algo dqn --env SpaceInvadersNoFrameskip-v4 -f logs/ -orga ItchyB ``` ## Hyperparameters ```python OrderedDict([('batch_size', 32), ('buffer_size', 100000), ('env_wrapper', ['stable_baselines3.common.atari_wrappers.AtariWrapper']), ('exploration_final_eps', 0.01), ('exploration_fraction', 0.1), ('frame_stack', 4), ('gradient_steps', 1), ('learning_rate', 0.0001), ('learning_starts', 100000), ('n_timesteps', 1000.0), ('optimize_memory_usage', False), ('policy', 'CnnPolicy'), ('target_update_interval', 1000), ('train_freq', 4), ('normalize', False)]) ```
2,681
[ [ -0.0406494140625, -0.03692626953125, 0.0219268798828125, 0.02545166015625, -0.009796142578125, -0.017333984375, 0.01224517822265625, -0.01348876953125, 0.01506805419921875, 0.02447509765625, -0.07000732421875, -0.0355224609375, -0.0270538330078125, -0.005256...
Phoshco/ADHDvsN
2023-04-09T16:57:14.000Z
[ "transformers", "pytorch", "bert", "text-classification", "generated_from_trainer", "license:apache-2.0", "endpoints_compatible", "region:us" ]
text-classification
Phoshco
null
null
Phoshco/ADHDvsN
0
2
transformers
2023-04-09T15:45:50
--- license: apache-2.0 tags: - generated_from_trainer metrics: - f1 - accuracy model-index: - name: ADHDvsN 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. --> # ADHDvsN This model is a fine-tuned version of [bert-base-cased](https://huggingface.co/bert-base-cased) on the None dataset. It achieves the following results on the evaluation set: - Loss: 0.7460 - F1: 0.684 - Roc Auc: 0.6836 - Accuracy: 0.684 ## 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: 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 | F1 | Roc Auc | Accuracy | |:-------------:|:-----:|:----:|:---------------:|:------:|:-------:|:--------:| | 0.6333 | 1.0 | 875 | 0.6383 | 0.6368 | 0.6321 | 0.6368 | | 0.591 | 2.0 | 1750 | 0.6384 | 0.6925 | 0.6926 | 0.6925 | | 0.5103 | 3.0 | 2625 | 0.6349 | 0.6827 | 0.6855 | 0.6827 | | 0.4122 | 4.0 | 3500 | 0.6424 | 0.668 | 0.6658 | 0.668 | | 0.3287 | 5.0 | 4375 | 0.7460 | 0.684 | 0.6836 | 0.684 | ### Framework versions - Transformers 4.27.3 - Pytorch 1.13.0 - Datasets 2.1.0 - Tokenizers 0.13.2
1,711
[ [ -0.040740966796875, -0.05078125, 0.00978851318359375, 0.001953125, -0.0177764892578125, -0.03216552734375, -0.004497528076171875, -0.01045989990234375, 0.0239715576171875, 0.0256805419921875, -0.063232421875, -0.049530029296875, -0.047637939453125, -0.022689...
xhorvat9/LTR_BERT_512_noTSD
2023-04-09T17:37:03.000Z
[ "transformers", "pytorch", "tensorboard", "bert", "text-classification", "generated_from_trainer", "endpoints_compatible", "region:us" ]
text-classification
xhorvat9
null
null
xhorvat9/LTR_BERT_512_noTSD
0
2
transformers
2023-04-09T15:59:42
--- tags: - generated_from_trainer metrics: - accuracy - precision - recall - f1 model-index: - name: output 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. --> # output This model is a fine-tuned version of [zhihan1996/DNA_bert_6](https://huggingface.co/zhihan1996/DNA_bert_6) on an unknown dataset. It achieves the following results on the evaluation set: - Loss: 0.3879 - Accuracy: 0.8612 - Precision: 0.9154 - Recall: 0.8240 - F1: 0.8673 ## 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: 16 - eval_batch_size: 16 - seed: 0 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 - lr_scheduler_type: linear - num_epochs: 3 ### Training results | Training Loss | Epoch | Step | Validation Loss | Accuracy | Precision | Recall | F1 | |:-------------:|:-----:|:----:|:---------------:|:--------:|:---------:|:------:|:------:| | 0.4958 | 0.46 | 500 | 0.4749 | 0.7831 | 0.7303 | 0.9606 | 0.8298 | | 0.3928 | 0.93 | 1000 | 0.4086 | 0.8207 | 0.7717 | 0.9574 | 0.8546 | | 0.3319 | 1.39 | 1500 | 0.3467 | 0.8635 | 0.8664 | 0.8891 | 0.8776 | | 0.3036 | 1.85 | 2000 | 0.3176 | 0.8702 | 0.8717 | 0.8960 | 0.8836 | | 0.2383 | 2.31 | 2500 | 0.3403 | 0.8707 | 0.8901 | 0.8728 | 0.8814 | | 0.2189 | 2.78 | 3000 | 0.3879 | 0.8612 | 0.9154 | 0.8240 | 0.8673 | ### Framework versions - Transformers 4.27.4 - Pytorch 2.0.0+cu118 - Tokenizers 0.13.3
1,901
[ [ -0.04217529296875, -0.039459228515625, 0.01263427734375, -0.00362396240234375, -0.0245513916015625, -0.025665283203125, -0.0086517333984375, -0.0153350830078125, 0.0206146240234375, 0.016571044921875, -0.060455322265625, -0.045013427734375, -0.046295166015625, ...
JaviBJ/ppo-SnowballTarget
2023-04-09T17:23:07.000Z
[ "ml-agents", "tensorboard", "onnx", "SnowballTarget", "deep-reinforcement-learning", "reinforcement-learning", "ML-Agents-SnowballTarget", "region:us" ]
reinforcement-learning
JaviBJ
null
null
JaviBJ/ppo-SnowballTarget
0
2
ml-agents
2023-04-09T17:23:02
--- library_name: ml-agents tags: - SnowballTarget - deep-reinforcement-learning - reinforcement-learning - ML-Agents-SnowballTarget --- # **ppo** Agent playing **SnowballTarget** This is a trained model of a **ppo** agent playing **SnowballTarget** using the [Unity ML-Agents Library](https://github.com/Unity-Technologies/ml-agents). ## Usage (with ML-Agents) The Documentation: https://github.com/huggingface/ml-agents#get-started We wrote a complete tutorial to learn to train your first agent using ML-Agents and publish it to the Hub: ### Resume the training ``` mlagents-learn <your_configuration_file_path.yaml> --run-id=<run_id> --resume ``` ### Watch your Agent play You can watch your agent **playing directly in your browser:**. 1. Go to https://huggingface.co/spaces/unity/ML-Agents-SnowballTarget 2. Step 1: Find your model_id: JaviBJ/ppo-SnowballTarget 3. Step 2: Select your *.nn /*.onnx file 4. Click on Watch the agent play 👀
985
[ [ -0.0164642333984375, -0.02783203125, 0.007312774658203125, 0.0165863037109375, -0.0221710205078125, 0.0170440673828125, 0.021820068359375, -0.00634765625, 0.0258331298828125, 0.03900146484375, -0.053070068359375, -0.05596923828125, -0.041351318359375, -0.017...
FCameCode/BERT_model
2023-04-10T11:25:48.000Z
[ "transformers", "pytorch", "tensorboard", "bert", "text-classification", "generated_from_trainer", "license:apache-2.0", "endpoints_compatible", "region:us" ]
text-classification
FCameCode
null
null
FCameCode/BERT_model
0
2
transformers
2023-04-09T17:37:10
--- license: apache-2.0 tags: - generated_from_trainer metrics: - accuracy model-index: - name: BERT_model 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. --> # BERT_model This model is a fine-tuned version of [bert-base-uncased](https://huggingface.co/bert-base-uncased) on the None dataset. It achieves the following results on the evaluation set: - Loss: 0.1260 - Accuracy: 0.9679 ## 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: 3 ### Training results | Training Loss | Epoch | Step | Validation Loss | Accuracy | |:-------------:|:-----:|:----:|:---------------:|:--------:| | 0.0934 | 1.0 | 1995 | 0.0993 | 0.9683 | | 0.0575 | 2.0 | 3990 | 0.1079 | 0.9695 | | 0.033 | 3.0 | 5985 | 0.1260 | 0.9679 | ### Framework versions - Transformers 4.27.4 - Pytorch 2.0.0+cu118 - Datasets 2.11.0 - Tokenizers 0.13.3
1,436
[ [ -0.03204345703125, -0.045928955078125, 0.01505279541015625, 0.011627197265625, -0.0277252197265625, -0.03680419921875, -0.0172271728515625, -0.02215576171875, 0.00977325439453125, 0.0238037109375, -0.054595947265625, -0.046356201171875, -0.047882080078125, -...
OtherBrian/distilbert-base-uncased-finetuned-emotion
2023-04-10T15:26:43.000Z
[ "transformers", "pytorch", "tensorboard", "distilbert", "text-classification", "generated_from_trainer", "dataset:emotion", "license:apache-2.0", "model-index", "endpoints_compatible", "region:us" ]
text-classification
OtherBrian
null
null
OtherBrian/distilbert-base-uncased-finetuned-emotion
0
2
transformers
2023-04-09T19:15:12
--- 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 config: split split: validation args: split metrics: - name: Accuracy type: accuracy value: 0.924 - name: F1 type: f1 value: 0.9240896354671038 --- <!-- 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: 0.2328 - Accuracy: 0.924 - F1: 0.9241 ## 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 | |:-------------:|:-----:|:----:|:---------------:|:--------:|:------:| | 0.8867 | 1.0 | 250 | 0.3406 | 0.9025 | 0.8973 | | 0.2671 | 2.0 | 500 | 0.2328 | 0.924 | 0.9241 | ### Framework versions - Transformers 4.27.4 - Pytorch 2.0.0+cu118 - Datasets 2.11.0 - Tokenizers 0.13.3
1,846
[ [ -0.0380859375, -0.042144775390625, 0.0158538818359375, 0.022003173828125, -0.026641845703125, -0.0194854736328125, -0.01329803466796875, -0.0088348388671875, 0.01056671142578125, 0.00878143310546875, -0.0565185546875, -0.051788330078125, -0.0596923828125, -0...
ValenHumano/roberta-base-bne-finetuned-amazon_reviews_multi
2023-04-09T22:13:02.000Z
[ "transformers", "pytorch", "tensorboard", "roberta", "text-classification", "generated_from_trainer", "dataset:amazon_reviews_multi", "license:apache-2.0", "model-index", "endpoints_compatible", "region:us" ]
text-classification
ValenHumano
null
null
ValenHumano/roberta-base-bne-finetuned-amazon_reviews_multi
1
2
transformers
2023-04-09T21:47:53
--- license: apache-2.0 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 config: es split: validation args: es metrics: - name: Accuracy type: accuracy value: 0.933 --- <!-- 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: 0.2233 - Accuracy: 0.933 ## 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 | |:-------------:|:-----:|:----:|:---------------:|:--------:| | 0.1956 | 1.0 | 1250 | 0.1798 | 0.9323 | | 0.107 | 2.0 | 2500 | 0.2233 | 0.933 | ### Framework versions - Transformers 4.27.4 - Pytorch 2.0.0+cu118 - Datasets 2.11.0 - Tokenizers 0.13.3
1,792
[ [ -0.039459228515625, -0.048736572265625, 0.0093994140625, 0.0140380859375, -0.027496337890625, -0.0303497314453125, -0.016845703125, -0.01898193359375, 0.0080413818359375, 0.028045654296875, -0.05072021484375, -0.045562744140625, -0.05316162109375, -0.0096282...
JyaouShingan/distilbert-base-uncased-local-finetuned-emotion
2023-04-12T03:51:49.000Z
[ "transformers", "pytorch", "distilbert", "text-classification", "generated_from_trainer", "dataset:emotion", "license:apache-2.0", "model-index", "endpoints_compatible", "region:us" ]
text-classification
JyaouShingan
null
null
JyaouShingan/distilbert-base-uncased-local-finetuned-emotion
0
2
transformers
2023-04-10T01:09:54
--- license: apache-2.0 tags: - generated_from_trainer datasets: - emotion metrics: - accuracy - f1 model-index: - name: distilbert-base-uncased-local-finetuned-emotion results: - task: name: Text Classification type: text-classification dataset: name: emotion type: emotion config: split split: validation args: split metrics: - name: Accuracy type: accuracy value: 0.9265 - name: F1 type: f1 value: 0.9264142965360822 --- <!-- 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-local-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: 0.2191 - Accuracy: 0.9265 - F1: 0.9264 ## 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 | |:-------------:|:-----:|:----:|:---------------:|:--------:|:------:| | No log | 1.0 | 250 | 0.3158 | 0.905 | 0.9026 | | No log | 2.0 | 500 | 0.2191 | 0.9265 | 0.9264 | ### Framework versions - Transformers 4.27.4 - Pytorch 2.0.0 - Datasets 2.11.0 - Tokenizers 0.13.2
1,854
[ [ -0.03680419921875, -0.042755126953125, 0.013458251953125, 0.0204620361328125, -0.024505615234375, -0.0205078125, -0.01488494873046875, -0.0124053955078125, 0.011993408203125, 0.0128021240234375, -0.0531005859375, -0.05230712890625, -0.054168701171875, -0.006...
Phoshco/allvsN
2023-04-10T06:02:33.000Z
[ "transformers", "pytorch", "bert", "text-classification", "generated_from_trainer", "license:apache-2.0", "endpoints_compatible", "region:us" ]
text-classification
Phoshco
null
null
Phoshco/allvsN
0
2
transformers
2023-04-10T03:43:36
--- license: apache-2.0 tags: - generated_from_trainer metrics: - f1 - accuracy model-index: - name: allvsN 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. --> # allvsN This model is a fine-tuned version of [bert-base-cased](https://huggingface.co/bert-base-cased) on the None dataset. It achieves the following results on the evaluation set: - Loss: 2.5592 - F1: 0.3265 - Accuracy: 0.3265 ## 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: 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 | F1 | Accuracy | |:-------------:|:-----:|:----:|:---------------:|:------:|:--------:| | 1.7706 | 1.0 | 1750 | 1.7077 | 0.3169 | 0.3169 | | 1.621 | 2.0 | 3500 | 1.6943 | 0.3396 | 0.3396 | | 1.3775 | 3.0 | 5250 | 1.7806 | 0.3458 | 0.3458 | | 0.9342 | 4.0 | 7000 | 2.0859 | 0.3406 | 0.3406 | | 0.5596 | 5.0 | 8750 | 2.5592 | 0.3265 | 0.3265 | ### Framework versions - Transformers 4.27.3 - Pytorch 1.13.0 - Datasets 2.1.0 - Tokenizers 0.13.2
1,623
[ [ -0.034423828125, -0.03997802734375, 0.01117706298828125, 0.012298583984375, -0.025909423828125, -0.036529541015625, -0.01027679443359375, -0.0105743408203125, 0.015869140625, 0.027313232421875, -0.05810546875, -0.051177978515625, -0.04510498046875, -0.019821...
rwang5688/distilbert-base-uncased-finetuned-sst2-pt
2023-09-25T06:42:27.000Z
[ "transformers", "pytorch", "tensorboard", "distilbert", "text-classification", "generated_from_trainer", "dataset:glue", "license:apache-2.0", "model-index", "endpoints_compatible", "region:us" ]
text-classification
rwang5688
null
null
rwang5688/distilbert-base-uncased-finetuned-sst2-pt
1
2
transformers
2023-04-10T04:35:09
--- license: apache-2.0 tags: - generated_from_trainer datasets: - glue metrics: - accuracy model-index: - name: distilbert-base-uncased-finetuned-sst2-pt results: - task: name: Text Classification type: text-classification dataset: name: glue type: glue config: sst2 split: validation args: sst2 metrics: - name: Accuracy type: accuracy value: 0.9071100917431193 --- <!-- 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-sst2-pt 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: 0.4661 - Accuracy: 0.9071 ## 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 | |:-------------:|:-----:|:-----:|:---------------:|:--------:| | 0.1863 | 1.0 | 4210 | 0.3161 | 0.8991 | | 0.1237 | 2.0 | 8420 | 0.3776 | 0.8956 | | 0.0997 | 3.0 | 12630 | 0.3770 | 0.9025 | | 0.0609 | 4.0 | 16840 | 0.4661 | 0.9071 | | 0.0376 | 5.0 | 21050 | 0.5535 | 0.9014 | ### Framework versions - Transformers 4.29.2 - Pytorch 2.0.1+cu118 - Datasets 2.12.0 - Tokenizers 0.13.2
1,923
[ [ -0.0213775634765625, -0.04766845703125, 0.0142364501953125, 0.013153076171875, -0.029632568359375, -0.01517486572265625, -0.0092620849609375, -0.001857757568359375, 0.0059051513671875, 0.0123291015625, -0.04718017578125, -0.038909912109375, -0.06292724609375, ...
Phoshco/bipolarvsN
2023-04-10T09:02:28.000Z
[ "transformers", "pytorch", "bert", "text-classification", "generated_from_trainer", "license:apache-2.0", "endpoints_compatible", "region:us" ]
text-classification
Phoshco
null
null
Phoshco/bipolarvsN
0
2
transformers
2023-04-10T07:44:01
--- license: apache-2.0 tags: - generated_from_trainer metrics: - f1 - accuracy model-index: - name: bipolarvsN 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. --> # bipolarvsN This model is a fine-tuned version of [bert-base-cased](https://huggingface.co/bert-base-cased) on the None dataset. It achieves the following results on the evaluation set: - Loss: 0.7437 - F1: 0.7833 - Roc Auc: 0.7818 - Accuracy: 0.7833 ## 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: 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 | F1 | Roc Auc | Accuracy | |:-------------:|:-----:|:----:|:---------------:|:------:|:-------:|:--------:| | 0.5251 | 1.0 | 875 | 0.5404 | 0.736 | 0.7299 | 0.736 | | 0.4396 | 2.0 | 1750 | 0.4694 | 0.7974 | 0.7966 | 0.7974 | | 0.373 | 3.0 | 2625 | 0.5041 | 0.797 | 0.7963 | 0.797 | | 0.2828 | 4.0 | 3500 | 0.6178 | 0.7939 | 0.7931 | 0.7939 | | 0.2147 | 5.0 | 4375 | 0.7437 | 0.7833 | 0.7818 | 0.7833 | ### Framework versions - Transformers 4.27.3 - Pytorch 1.13.0 - Datasets 2.1.0 - Tokenizers 0.13.2
1,719
[ [ -0.03875732421875, -0.042877197265625, 0.01053619384765625, 0.0131072998046875, -0.029998779296875, -0.0225982666015625, -0.00484466552734375, -0.003910064697265625, 0.0208587646484375, 0.03509521484375, -0.060150146484375, -0.058563232421875, -0.048309326171875...
Amite5h/TextClassificationmulticlass
2023-04-10T08:52:58.000Z
[ "transformers", "tf", "distilbert", "text-classification", "generated_from_keras_callback", "license:apache-2.0", "endpoints_compatible", "region:us" ]
text-classification
Amite5h
null
null
Amite5h/TextClassificationmulticlass
1
2
transformers
2023-04-10T08:43:47
--- license: apache-2.0 tags: - generated_from_keras_callback model-index: - name: TextClassificationmulticlass results: [] --- <!-- This model card has been generated automatically according to the information Keras had access to. You should probably proofread and complete it, then remove this comment. --> # TextClassificationmulticlass 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: ## 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: - optimizer: {'name': 'Adam', 'weight_decay': None, 'clipnorm': None, 'global_clipnorm': None, 'clipvalue': None, 'use_ema': False, 'ema_momentum': 0.99, 'ema_overwrite_frequency': None, 'jit_compile': True, 'is_legacy_optimizer': False, 'learning_rate': 5e-05, 'beta_1': 0.9, 'beta_2': 0.999, 'epsilon': 1e-07, 'amsgrad': False} - training_precision: float32 ### Training results ### Framework versions - Transformers 4.28.0.dev0 - TensorFlow 2.12.0 - Tokenizers 0.13.3
1,287
[ [ -0.038909912109375, -0.0521240234375, 0.024261474609375, 0.0029773712158203125, -0.0411376953125, -0.0109100341796875, -0.0156097412109375, -0.01605224609375, 0.006259918212890625, 0.008026123046875, -0.044219970703125, -0.052642822265625, -0.0648193359375, ...
chanelcolgate/vit-base-patch16-224-chest-x-ray
2023-04-10T09:08:15.000Z
[ "transformers", "pytorch", "vit", "image-classification", "generated_from_trainer", "dataset:chest-xray-classification", "license:apache-2.0", "autotrain_compatible", "endpoints_compatible", "region:us" ]
image-classification
chanelcolgate
null
null
chanelcolgate/vit-base-patch16-224-chest-x-ray
0
2
transformers
2023-04-10T08:45:56
--- license: apache-2.0 tags: - generated_from_trainer datasets: - chest-xray-classification model-index: - name: vit-base-patch16-224-chest-x-ray 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. --> # vit-base-patch16-224-chest-x-ray This model is a fine-tuned version of [google/vit-base-patch16-224](https://huggingface.co/google/vit-base-patch16-224) on the chest-xray-classification 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: 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 ### Framework versions - Transformers 4.27.4 - Pytorch 2.0.0+cu118 - Datasets 2.11.0 - Tokenizers 0.13.3
1,114
[ [ -0.028289794921875, -0.037353515625, 0.01309967041015625, -0.00955963134765625, -0.04412841796875, -0.0261993408203125, 0.0096435546875, -0.0160675048828125, 0.0113372802734375, 0.03662109375, -0.050567626953125, -0.0406494140625, -0.050323486328125, -0.0151...
GhifSmile/xlm-roberta-base-uncased-PINA
2023-04-10T10:11:16.000Z
[ "transformers", "pytorch", "tensorboard", "xlm-roberta", "text-classification", "generated_from_trainer", "license:mit", "endpoints_compatible", "region:us" ]
text-classification
GhifSmile
null
null
GhifSmile/xlm-roberta-base-uncased-PINA
0
2
transformers
2023-04-10T08:54:45
--- license: mit tags: - generated_from_trainer metrics: - accuracy - precision - recall model-index: - name: xlm-roberta-base-uncased-PINA 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. --> # xlm-roberta-base-uncased-PINA This model is a fine-tuned version of [xlm-roberta-base](https://huggingface.co/xlm-roberta-base) on the None dataset. It achieves the following results on the evaluation set: - Loss: 1.0862 - Accuracy: 0.7553 - Precision: 0.5016 - Recall: 0.4522 ## 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: 10 ### Training results | Training Loss | Epoch | Step | Validation Loss | Accuracy | Precision | Recall | |:-------------:|:-----:|:----:|:---------------:|:--------:|:---------:|:------:| | 2.7204 | 1.0 | 234 | 2.4959 | 0.4220 | 0.0124 | 0.0269 | | 2.2553 | 2.0 | 468 | 1.9819 | 0.5 | 0.0498 | 0.0802 | | 1.9593 | 3.0 | 702 | 1.7527 | 0.5513 | 0.1222 | 0.1377 | | 1.6947 | 4.0 | 936 | 1.5375 | 0.6325 | 0.2466 | 0.2480 | | 1.4593 | 5.0 | 1170 | 1.3773 | 0.6848 | 0.4074 | 0.3414 | | 1.2381 | 6.0 | 1404 | 1.2560 | 0.7094 | 0.4273 | 0.3638 | | 1.0986 | 7.0 | 1638 | 1.1813 | 0.7286 | 0.4396 | 0.4033 | | 0.9817 | 8.0 | 1872 | 1.1668 | 0.7361 | 0.4824 | 0.4345 | | 0.8894 | 9.0 | 2106 | 1.1054 | 0.7521 | 0.5155 | 0.4461 | | 0.8518 | 10.0 | 2340 | 1.0862 | 0.7553 | 0.5016 | 0.4522 | ### Framework versions - Transformers 4.27.4 - Pytorch 2.0.0+cu118 - Datasets 2.11.0 - Tokenizers 0.13.3
2,210
[ [ -0.03680419921875, -0.042144775390625, 0.0216064453125, 0.0018157958984375, -0.0141143798828125, -0.01934814453125, -0.005767822265625, -0.0115814208984375, 0.0234832763671875, 0.031982421875, -0.04998779296875, -0.052520751953125, -0.053131103515625, -0.011...
Seungjun/articleGeneratorV1.0
2023-04-10T10:19:56.000Z
[ "transformers", "tf", "t5", "text2text-generation", "generated_from_keras_callback", "license:apache-2.0", "autotrain_compatible", "endpoints_compatible", "text-generation-inference", "region:us" ]
text2text-generation
Seungjun
null
null
Seungjun/articleGeneratorV1.0
1
2
transformers
2023-04-10T09:02:51
--- license: apache-2.0 tags: - generated_from_keras_callback model-index: - name: articleGeneratorV1.0 results: [] --- <!-- This model card has been generated automatically according to the information Keras had access to. You should probably proofread and complete it, then remove this comment. --> # What does model do and how to use it Just provide an title to the model and it will generate a whole article about it. ```python # Install transformers library !pip install transformers ``` ```python # Load tokenizer and model from transformers import AutoTokenizer, AutoModelForSeq2SeqLM, TFAutoModelForSeq2SeqLM model_name = "Seungjun/articleGeneratorV1.0" tokenizer = AutoTokenizer.from_pretrained("t5-small") model = TFAutoModelForSeq2SeqLM.from_pretrained(model_name) ``` ```python # Get the article for a given title from transformers import pipeline summarizer = pipeline("summarization", model=model, tokenizer=tokenizer, framework="tf") summarizer( "Steve Jobs", # title min_length=500, max_length=1024, ) ``` Result: # Current limitation of the model It generate aot of lies. 99% of the word generated by this model is not true. # articleGeneratorV1.0 This model is a fine-tuned version of [t5-small](https://huggingface.co/t5-small) on an unknown dataset. It achieves the following results on the evaluation set: - Train Loss: 3.9568 - Validation Loss: 3.6096 - Train Rougel: tf.Tensor(0.08172019, shape=(), dtype=float32) - Epoch: 4 ## 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: - optimizer: {'name': 'Adam', 'weight_decay': None, 'clipnorm': None, 'global_clipnorm': None, 'clipvalue': None, 'use_ema': False, 'ema_momentum': 0.99, 'ema_overwrite_frequency': None, 'jit_compile': True, 'is_legacy_optimizer': False, 'learning_rate': 2e-05, 'beta_1': 0.9, 'beta_2': 0.999, 'epsilon': 1e-07, 'amsgrad': False} - training_precision: float32 ### Training results | Train Loss | Validation Loss | Train Rougel | Epoch | |:----------:|:---------------:|:-----------------------------------------------:|:-----:| | 4.9218 | 4.0315 | tf.Tensor(0.08038119, shape=(), dtype=float32) | 0 | | 4.2887 | 3.8366 | tf.Tensor(0.08103053, shape=(), dtype=float32) | 1 | | 4.1269 | 3.7328 | tf.Tensor(0.081041485, shape=(), dtype=float32) | 2 | | 4.0276 | 3.6614 | tf.Tensor(0.081364945, shape=(), dtype=float32) | 3 | | 3.9568 | 3.6096 | tf.Tensor(0.08172019, shape=(), dtype=float32) | 4 | ### Framework versions - Transformers 4.27.4 - TensorFlow 2.12.0 - Datasets 2.11.0 - Tokenizers 0.13.3
2,875
[ [ -0.031494140625, -0.0384521484375, 0.02978515625, -0.0003437995910644531, -0.0205841064453125, -0.022705078125, -0.0007805824279785156, -0.01409912109375, 0.01947021484375, 0.0169830322265625, -0.04400634765625, -0.0545654296875, -0.05584716796875, -0.007942...
Phoshco/depressionvsN
2023-04-10T10:24:59.000Z
[ "transformers", "pytorch", "bert", "text-classification", "generated_from_trainer", "license:apache-2.0", "endpoints_compatible", "region:us" ]
text-classification
Phoshco
null
null
Phoshco/depressionvsN
0
2
transformers
2023-04-10T09:06:06
--- license: apache-2.0 tags: - generated_from_trainer metrics: - f1 - accuracy model-index: - name: depressionvsN 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. --> # depressionvsN This model is a fine-tuned version of [bert-base-cased](https://huggingface.co/bert-base-cased) on the None dataset. It achieves the following results on the evaluation set: - Loss: 1.1005 - F1: 0.6615 - Roc Auc: 0.6610 - Accuracy: 0.6615 ## 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: 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 | F1 | Roc Auc | Accuracy | |:-------------:|:-----:|:----:|:---------------:|:------:|:-------:|:--------:| | 0.6688 | 1.0 | 875 | 0.6239 | 0.6552 | 0.6546 | 0.6552 | | 0.5832 | 2.0 | 1750 | 0.5966 | 0.6786 | 0.6789 | 0.6786 | | 0.4778 | 3.0 | 2625 | 0.6958 | 0.6791 | 0.6795 | 0.6791 | | 0.3487 | 4.0 | 3500 | 0.7418 | 0.6637 | 0.6617 | 0.6637 | | 0.2266 | 5.0 | 4375 | 1.1005 | 0.6615 | 0.6610 | 0.6615 | ### Framework versions - Transformers 4.27.3 - Pytorch 1.13.0 - Datasets 2.1.0 - Tokenizers 0.13.2
1,725
[ [ -0.039581298828125, -0.0439453125, 0.0132293701171875, 0.011993408203125, -0.0229644775390625, -0.031463623046875, -0.00940704345703125, -0.0099334716796875, 0.0216217041015625, 0.0302581787109375, -0.06268310546875, -0.0540771484375, -0.047821044921875, -0....
sakethchalla/isl-nodel
2023-04-10T10:34:41.000Z
[ "transformers", "pytorch", "tensorboard", "vit", "image-classification", "generated_from_trainer", "dataset:imagefolder", "license:apache-2.0", "model-index", "autotrain_compatible", "endpoints_compatible", "region:us" ]
image-classification
sakethchalla
null
null
sakethchalla/isl-nodel
0
2
transformers
2023-04-10T09:51:28
--- license: apache-2.0 tags: - generated_from_trainer datasets: - imagefolder metrics: - accuracy model-index: - name: isl-nodel results: - task: name: Image Classification type: image-classification dataset: name: imagefolder type: imagefolder config: default split: train args: default metrics: - name: Accuracy type: accuracy value: 0.7540407589599438 --- <!-- 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. --> # isl-nodel This model is a fine-tuned version of [google/vit-base-patch16-224-in21k](https://huggingface.co/google/vit-base-patch16-224-in21k) on the imagefolder dataset. It achieves the following results on the evaluation set: - Loss: 0.9554 - Accuracy: 0.7540 ## 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: 16 - eval_batch_size: 16 - seed: 42 - gradient_accumulation_steps: 4 - total_train_batch_size: 64 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 - lr_scheduler_type: linear - lr_scheduler_warmup_ratio: 0.1 - num_epochs: 10 ### Training results | Training Loss | Epoch | Step | Validation Loss | Accuracy | |:-------------:|:-----:|:----:|:---------------:|:--------:| | 2.6213 | 1.0 | 89 | 2.3886 | 0.6128 | | 1.66 | 2.0 | 178 | 1.5769 | 0.7119 | | 1.3588 | 3.0 | 267 | 1.3264 | 0.7358 | | 1.1062 | 4.0 | 356 | 1.1833 | 0.7386 | | 1.1883 | 5.0 | 445 | 1.1025 | 0.7442 | | 1.159 | 6.0 | 534 | 1.0324 | 0.7505 | | 0.9934 | 7.0 | 623 | 0.9626 | 0.7674 | | 0.8885 | 8.0 | 712 | 1.0080 | 0.7435 | | 0.9325 | 9.0 | 801 | 0.9395 | 0.7681 | | 0.9254 | 10.0 | 890 | 0.9554 | 0.7540 | ### Framework versions - Transformers 4.27.4 - Pytorch 2.0.0+cu118 - Datasets 2.11.0 - Tokenizers 0.13.3
2,309
[ [ -0.0318603515625, -0.04571533203125, 0.007221221923828125, 0.006542205810546875, -0.0262908935546875, -0.02020263671875, -0.0015058517456054688, -0.017120361328125, 0.01224517822265625, 0.0220794677734375, -0.05303955078125, -0.0513916015625, -0.051239013671875,...
Phoshco/EDAnonymousvsN
2023-04-10T11:47:43.000Z
[ "transformers", "pytorch", "bert", "text-classification", "generated_from_trainer", "license:apache-2.0", "endpoints_compatible", "region:us" ]
text-classification
Phoshco
null
null
Phoshco/EDAnonymousvsN
0
2
transformers
2023-04-10T10:28:33
--- license: apache-2.0 tags: - generated_from_trainer metrics: - f1 - accuracy model-index: - name: EDAnonymousvsN 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. --> # EDAnonymousvsN This model is a fine-tuned version of [bert-base-cased](https://huggingface.co/bert-base-cased) on the None dataset. It achieves the following results on the evaluation set: - Loss: 0.4876 - F1: 0.8914 - Roc Auc: 0.8899 - Accuracy: 0.8914 ## 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: 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 | F1 | Roc Auc | Accuracy | |:-------------:|:-----:|:----:|:---------------:|:------:|:-------:|:--------:| | 0.3665 | 1.0 | 875 | 0.3116 | 0.889 | 0.8858 | 0.889 | | 0.253 | 2.0 | 1750 | 0.2832 | 0.8884 | 0.8866 | 0.8884 | | 0.2082 | 3.0 | 2625 | 0.3573 | 0.8934 | 0.8915 | 0.8934 | | 0.1422 | 4.0 | 3500 | 0.4506 | 0.8932 | 0.8926 | 0.8932 | | 0.0953 | 5.0 | 4375 | 0.4876 | 0.8914 | 0.8899 | 0.8914 | ### Framework versions - Transformers 4.27.3 - Pytorch 1.13.0 - Datasets 2.1.0 - Tokenizers 0.13.2
1,727
[ [ -0.037353515625, -0.045928955078125, 0.01233673095703125, 0.00923919677734375, -0.0251312255859375, -0.03302001953125, -0.01062774658203125, -0.0124969482421875, 0.0245513916015625, 0.0307159423828125, -0.061798095703125, -0.0556640625, -0.047119140625, -0.0...
mnavas/roberta-finetuned-solvencia-v1
2023-04-12T13:27:52.000Z
[ "transformers", "pytorch", "tensorboard", "roberta", "text-classification", "generated_from_trainer", "license:mit", "endpoints_compatible", "region:us" ]
text-classification
mnavas
null
null
mnavas/roberta-finetuned-solvencia-v1
0
2
transformers
2023-04-10T10:49:56
--- license: mit tags: - generated_from_trainer metrics: - accuracy - f1 - precision - recall model-index: - name: roberta-finetuned-solvencia-v1 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. --> # roberta-finetuned-solvencia-v1 This model is a fine-tuned version of [xlm-roberta-base](https://huggingface.co/xlm-roberta-base) on the None dataset. It achieves the following results on the evaluation set: - Loss: 0.4141 - Accuracy: 0.8919 - F1: 0.8919 - Precision: 0.8919 - Recall: 0.8919 ## 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: 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: 10 ### Training results | Training Loss | Epoch | Step | Validation Loss | Accuracy | F1 | Precision | Recall | |:-------------:|:-----:|:----:|:---------------:|:--------:|:------:|:---------:|:------:| | No log | 1.0 | 333 | 0.3781 | 0.8544 | 0.8544 | 0.8544 | 0.8544 | | 0.4429 | 2.0 | 666 | 0.3295 | 0.8679 | 0.8679 | 0.8679 | 0.8679 | | 0.4429 | 3.0 | 999 | 0.3664 | 0.8784 | 0.8784 | 0.8784 | 0.8784 | | 0.3512 | 4.0 | 1332 | 0.4602 | 0.8649 | 0.8649 | 0.8649 | 0.8649 | | 0.2975 | 5.0 | 1665 | 0.4721 | 0.8889 | 0.8889 | 0.8889 | 0.8889 | | 0.2975 | 6.0 | 1998 | 0.4141 | 0.8919 | 0.8919 | 0.8919 | 0.8919 | | 0.2499 | 7.0 | 2331 | 0.4054 | 0.8889 | 0.8889 | 0.8889 | 0.8889 | | 0.2132 | 8.0 | 2664 | 0.4878 | 0.8829 | 0.8829 | 0.8829 | 0.8829 | | 0.2132 | 9.0 | 2997 | 0.4867 | 0.8904 | 0.8904 | 0.8904 | 0.8904 | | 0.1812 | 10.0 | 3330 | 0.5339 | 0.8889 | 0.8889 | 0.8889 | 0.8889 | ### Framework versions - Transformers 4.16.2 - Pytorch 1.9.1 - Datasets 1.18.4 - Tokenizers 0.11.6
2,330
[ [ -0.036529541015625, -0.0426025390625, 0.015380859375, 0.0016307830810546875, -0.0178985595703125, -0.017913818359375, -0.003856658935546875, -0.01021575927734375, 0.0266265869140625, 0.0318603515625, -0.058013916015625, -0.051025390625, -0.053009033203125, -...
Phoshco/ptsdvsN
2023-04-10T13:26:28.000Z
[ "transformers", "pytorch", "bert", "text-classification", "generated_from_trainer", "license:apache-2.0", "endpoints_compatible", "region:us" ]
text-classification
Phoshco
null
null
Phoshco/ptsdvsN
0
2
transformers
2023-04-10T12:08:05
--- license: apache-2.0 tags: - generated_from_trainer metrics: - f1 - accuracy model-index: - name: ptsdvsN 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. --> # ptsdvsN This model is a fine-tuned version of [bert-base-cased](https://huggingface.co/bert-base-cased) on the None dataset. It achieves the following results on the evaluation set: - Loss: 1.0050 - F1: 0.8051 - Roc Auc: 0.8042 - Accuracy: 0.8051 ## 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: 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 | F1 | Roc Auc | Accuracy | |:-------------:|:-----:|:----:|:---------------:|:------:|:-------:|:--------:| | 0.4907 | 1.0 | 875 | 0.4679 | 0.8161 | 0.8164 | 0.8161 | | 0.3568 | 2.0 | 1750 | 0.4654 | 0.8221 | 0.8225 | 0.8221 | | 0.2289 | 3.0 | 2625 | 0.7412 | 0.7843 | 0.7800 | 0.7843 | | 0.1246 | 4.0 | 3500 | 0.8720 | 0.8013 | 0.7995 | 0.8013 | | 0.0656 | 5.0 | 4375 | 1.0050 | 0.8051 | 0.8042 | 0.8051 | ### Framework versions - Transformers 4.27.3 - Pytorch 1.13.0 - Datasets 2.1.0 - Tokenizers 0.13.2
1,713
[ [ -0.037078857421875, -0.04510498046875, 0.01284027099609375, 0.010284423828125, -0.02587890625, -0.031494140625, -0.00720977783203125, -0.0105133056640625, 0.0198516845703125, 0.03106689453125, -0.060211181640625, -0.05322265625, -0.048858642578125, -0.016708...
justinsiow/PyramdisRND
2023-04-10T14:20:05.000Z
[ "ml-agents", "tensorboard", "onnx", "Pyramids", "deep-reinforcement-learning", "reinforcement-learning", "ML-Agents-Pyramids", "region:us" ]
reinforcement-learning
justinsiow
null
null
justinsiow/PyramdisRND
0
2
ml-agents
2023-04-10T14:18:06
--- library_name: ml-agents tags: - Pyramids - deep-reinforcement-learning - reinforcement-learning - ML-Agents-Pyramids --- # **ppo** Agent playing **Pyramids** This is a trained model of a **ppo** agent playing **Pyramids** using the [Unity ML-Agents Library](https://github.com/Unity-Technologies/ml-agents). ## Usage (with ML-Agents) The Documentation: https://github.com/huggingface/ml-agents#get-started We wrote a complete tutorial to learn to train your first agent using ML-Agents and publish it to the Hub: ### Resume the training ``` mlagents-learn <your_configuration_file_path.yaml> --run-id=<run_id> --resume ``` ### Watch your Agent play You can watch your agent **playing directly in your browser:**. 1. Go to https://huggingface.co/spaces/unity/ML-Agents-Pyramids 2. Step 1: Find your model_id: justinsiow/PyramdisRND 3. Step 2: Select your *.nn /*.onnx file 4. Click on Watch the agent play 👀
952
[ [ -0.0265960693359375, -0.020050048828125, 0.0011968612670898438, 0.026275634765625, -0.00969696044921875, 0.005889892578125, 0.0276336669921875, -0.0035953521728515625, 0.0347900390625, 0.036651611328125, -0.036895751953125, -0.050537109375, -0.036346435546875, ...
AyoubChLin/distilbert_cnn_news
2023-05-07T15:00:42.000Z
[ "transformers", "pytorch", "distilbert", "text-classification", "en", "dataset:AyoubChLin/CNN_News_Articles_2011-2022", "license:apache-2.0", "endpoints_compatible", "region:us" ]
text-classification
AyoubChLin
null
null
AyoubChLin/distilbert_cnn_news
1
2
transformers
2023-04-10T14:52:20
--- license: apache-2.0 datasets: - AyoubChLin/CNN_News_Articles_2011-2022 language: - en metrics: - accuracy pipeline_tag: text-classification widget: - text: money in the pocket - text: no one can win this cup in quatar.. - text: Health is an essential aspect of our lives that affects us physically, mentally, and emotionally. Maintaining good health requires us to make healthy lifestyle choices, including eating a balanced diet, getting regular exercise, and getting enough sleep. These habits can help reduce the risk of developing chronic diseases such as diabetes, heart disease, and cancer. --- ## DistilBertForSequenceClassification on CNN News Dataset This repository contains a fine-tuned DistilBert base model for sequence classification on the CNN News dataset. The model is able to classify news articles into one of six categories: business, entertainment, health, news, politics, and sport. The model was fine-tuned for four epochs achieving a training loss of 0.012900, a validation loss of 0.151663, - accuracy of 0.9607394366197183. - f1 : 0.962072 - precision : 0.961904 - recall : 0.962324 ### Model Description <!-- Provide a longer summary of what this model is. --> - **Developed by:** [CHERGUELAINE Ayoub](https://www.linkedin.com/in/ayoub-cherguelaine/) & [BOUBEKRI Faycal](https://www.linkedin.com/in/faycal-boubekri-832848199/) - **Shared by [optional]:** HuggingFace - **Model type:** Language model - **Language(s) (NLP):** en - **Finetuned from model [optional]:** [distilbert-base-uncased](https://huggingface.co/distilbert-base-uncased) ### Usage You can use this model with the Hugging Face Transformers library for a variety of natural language processing tasks, such as text classification, sentiment analysis, and more. Here's an example of how to use this model for text classification in Python: ``` python from transformers import AutoTokenizer, DistilBertForSequenceClassification model_name = "AyoubChLin/distilbert_cnn_news" tokenizer = AutoTokenizer.from_pretrained(model_name) model = TFAutoModelForSequenceClassification.from_pretrained(model_name) text = "This is a news article about politics." inputs = tokenizer(text, padding=True, truncation=True, return_tensors="pt") with torch.no_grad(): logits = model(**inputs).logits predicted_class_id = logits.argmax().item() ``` In this example, we first load the tokenizer and the model using their respective from_pretrained methods. We then encode a news article using the tokenizer, pass the inputs through the model, and extract the predicted label using the argmax function. Finally, we map the predicted label to its corresponding category using a list of labels. ### Contributors This model was fine-tuned by CHERGUELAINE Ayoub and BOUBEKRI Faycal.
2,781
[ [ -0.0245819091796875, -0.047637939453125, 0.00920867919921875, 0.01549530029296875, -0.0177001953125, 0.005886077880859375, -0.006946563720703125, -0.0172576904296875, -0.00508880615234375, 0.012420654296875, -0.028564453125, -0.049346923828125, -0.064208984375, ...
galkowskim/distilbert-base-uncased-finetuned-emotions
2023-04-10T16:13:01.000Z
[ "transformers", "pytorch", "tensorboard", "distilbert", "text-classification", "generated_from_trainer", "dataset:emotion", "license:apache-2.0", "model-index", "endpoints_compatible", "region:us" ]
text-classification
galkowskim
null
null
galkowskim/distilbert-base-uncased-finetuned-emotions
0
2
transformers
2023-04-10T15:45:41
--- license: apache-2.0 tags: - generated_from_trainer datasets: - emotion metrics: - accuracy - f1 model-index: - name: distilbert-base-uncased-finetuned-emotions results: - task: name: Text Classification type: text-classification dataset: name: emotion type: emotion config: split split: validation args: split metrics: - name: Accuracy type: accuracy value: 0.9255 - name: F1 type: f1 value: 0.9255657653416817 --- <!-- 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-emotions 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: 0.2190 - Accuracy: 0.9255 - F1: 0.9256 ## 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 | |:-------------:|:-----:|:----:|:---------------:|:--------:|:------:| | 0.8373 | 1.0 | 250 | 0.3222 | 0.903 | 0.8990 | | 0.2494 | 2.0 | 500 | 0.2190 | 0.9255 | 0.9256 | ### Framework versions - Transformers 4.27.4 - Pytorch 2.0.0+cu118 - Datasets 2.11.0 - Tokenizers 0.13.3
1,850
[ [ -0.039337158203125, -0.041473388671875, 0.01485443115234375, 0.021209716796875, -0.0269927978515625, -0.0194854736328125, -0.0135345458984375, -0.0081787109375, 0.007419586181640625, 0.007656097412109375, -0.057647705078125, -0.051605224609375, -0.05868530273437...
Svetlana0303/regression_Albert_1500
2023-04-10T15:54:44.000Z
[ "transformers", "tf", "albert", "text-classification", "generated_from_keras_callback", "license:apache-2.0", "endpoints_compatible", "region:us" ]
text-classification
Svetlana0303
null
null
Svetlana0303/regression_Albert_1500
0
2
transformers
2023-04-10T15:54:38
--- license: apache-2.0 tags: - generated_from_keras_callback model-index: - name: Regression_bert_1500 results: [] --- <!-- This model card has been generated automatically according to the information Keras had access to. You should probably proofread and complete it, then remove this comment. --> # Regression_bert_1500 This model is a fine-tuned version of [albert-base-v2](https://huggingface.co/albert-base-v2) on an unknown dataset. It achieves the following results on the evaluation set: - Train Loss: 0.3665 - Train Mae: 0.5651 - Train Mse: 0.4539 - Train R2-score: 0.5632 - Validation Loss: 0.3640 - Validation Mae: 0.6123 - Validation Mse: 0.4470 - Validation R2-score: 0.5765 - Epoch: 22 ## 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: - optimizer: {'name': 'Adam', 'weight_decay': None, 'clipnorm': None, 'global_clipnorm': None, 'clipvalue': None, 'use_ema': False, 'ema_momentum': 0.99, 'ema_overwrite_frequency': None, 'jit_compile': True, 'is_legacy_optimizer': False, 'learning_rate': 1e-04, 'beta_1': 0.9, 'beta_2': 0.999, 'epsilon': 1e-07, 'amsgrad': False} - training_precision: float32 ### Training results | Train Loss | Train Mae | Train Mse | Train R2-score | Validation Loss | Validation Mae | Validation Mse | Validation R2-score | Epoch | |:----------:|:---------:|:---------:|:--------------:|:---------------:|:--------------:|:--------------:|:-------------------:|:-----:| | 0.3911 | 0.5811 | 0.4875 | 0.5636 | 0.3808 | 0.6393 | 0.4778 | 0.4775 | 0 | | 0.3669 | 0.5644 | 0.4527 | 0.6196 | 0.3524 | 0.5673 | 0.4286 | 0.6944 | 1 | | 0.3652 | 0.5606 | 0.4457 | 0.6645 | 0.3711 | 0.6253 | 0.4600 | 0.5315 | 2 | | 0.3669 | 0.5642 | 0.4490 | 0.5194 | 0.3525 | 0.5695 | 0.4286 | 0.6901 | 3 | | 0.3693 | 0.5693 | 0.4580 | 0.6646 | 0.3558 | 0.5904 | 0.4329 | 0.6414 | 4 | | 0.3682 | 0.5633 | 0.4540 | 0.7464 | 0.3602 | 0.5255 | 0.4485 | 0.7509 | 5 | | 0.3712 | 0.5632 | 0.4527 | 0.6645 | 0.3650 | 0.6145 | 0.4489 | 0.5693 | 6 | | 0.3781 | 0.5720 | 0.4661 | 0.5801 | 0.3545 | 0.5849 | 0.4309 | 0.6553 | 7 | | 0.3659 | 0.5673 | 0.4564 | 0.1693 | 0.3723 | 0.6271 | 0.4621 | 0.5247 | 8 | | 0.3693 | 0.5642 | 0.4487 | 0.7048 | 0.3524 | 0.5641 | 0.4289 | 0.7006 | 9 | | 0.3656 | 0.5655 | 0.4495 | 0.6565 | 0.3575 | 0.5328 | 0.4425 | 0.7448 | 10 | | 0.3685 | 0.5632 | 0.4540 | 0.7202 | 0.3551 | 0.5878 | 0.4319 | 0.6482 | 11 | | 0.3702 | 0.5646 | 0.4543 | 0.7295 | 0.3528 | 0.5557 | 0.4306 | 0.7152 | 12 | | 0.3661 | 0.5615 | 0.4450 | 0.6631 | 0.3683 | 0.5240 | 0.4664 | 0.7592 | 13 | | 0.3835 | 0.5742 | 0.4757 | 0.7335 | 0.3531 | 0.5523 | 0.4316 | 0.7206 | 14 | | 0.3641 | 0.5628 | 0.4472 | 0.7325 | 0.3559 | 0.5909 | 0.4331 | 0.6399 | 15 | | 0.3764 | 0.5633 | 0.4566 | 0.7291 | 0.3549 | 0.5867 | 0.4315 | 0.6508 | 16 | | 0.3625 | 0.5594 | 0.4443 | 0.5555 | 0.3648 | 0.6141 | 0.4486 | 0.5707 | 17 | | 0.3816 | 0.5743 | 0.4693 | 0.6649 | 0.3559 | 0.5385 | 0.4385 | 0.7389 | 18 | | 0.3721 | 0.5721 | 0.4618 | 0.6791 | 0.3529 | 0.5745 | 0.4288 | 0.6795 | 19 | | 0.3711 | 0.5659 | 0.4586 | 0.2709 | 0.3610 | 0.5234 | 0.4505 | 0.7525 | 20 | | 0.3693 | 0.5641 | 0.4501 | 0.7400 | 0.3525 | 0.5607 | 0.4294 | 0.7068 | 21 | | 0.3665 | 0.5651 | 0.4539 | 0.5632 | 0.3640 | 0.6123 | 0.4470 | 0.5765 | 22 | ### Framework versions - Transformers 4.27.4 - TensorFlow 2.12.0 - Datasets 2.11.0 - Tokenizers 0.13.3
4,919
[ [ -0.049835205078125, -0.0455322265625, 0.0177154541015625, 0.0016050338745117188, -0.00238800048828125, 0.0010175704956054688, -0.0016937255859375, -0.005374908447265625, 0.051910400390625, 0.02447509765625, -0.048919677734375, -0.047576904296875, -0.049835205078...
Roguwan/DialoGPT-medium-rogu
2023-04-10T20:58:16.000Z
[ "transformers", "pytorch", "gpt2", "text-generation", "conversational", "license:mit", "endpoints_compatible", "text-generation-inference", "region:us" ]
conversational
Roguwan
null
null
Roguwan/DialoGPT-medium-rogu
0
2
transformers
2023-04-10T17:12:09
--- tags: - conversational license: mit --- ```python from transformers import AutoTokenizer, AutoModelWithLMHead tokenizer = AutoTokenizer.from_pretrained("r3dhummingbird/DialoGPT-medium-joshua") model = AutoModelWithLMHead.from_pretrained("r3dhummingbird/DialoGPT-medium-joshua") # Let's chat for 4 lines for step in range(4): # encode the new user input, add the eos_token and return a tensor in Pytorch new_user_input_ids = tokenizer.encode(input(">> User:") + tokenizer.eos_token, return_tensors='pt') # print(new_user_input_ids) # append the new user input tokens to the chat history bot_input_ids = torch.cat([chat_history_ids, new_user_input_ids], dim=-1) if step > 0 else new_user_input_ids # generated a response while limiting the total chat history to 1000 tokens, chat_history_ids = model.generate( bot_input_ids, max_length=200, pad_token_id=tokenizer.eos_token_id, no_repeat_ngram_size=3, do_sample=True, top_k=100, top_p=0.7, temperature=0.8 ) # pretty print last ouput tokens from bot print("JoshuaBot: {}".format(tokenizer.decode(chat_history_ids[:, bot_input_ids.shape[-1]:][0], skip_special_tokens=True))) ```
1,251
[ [ -0.0238494873046875, -0.044464111328125, -0.0007958412170410156, 0.0179595947265625, -0.0136260986328125, 0.00777435302734375, -0.003513336181640625, 0.00392913818359375, 0.0161590576171875, 0.01512908935546875, -0.048614501953125, -0.0193939208984375, -0.059997...
jmurphy97/dqn-SpaceInvadersNoFrameskip-v4
2023-04-10T20:52:16.000Z
[ "stable-baselines3", "SpaceInvadersNoFrameskip-v4", "deep-reinforcement-learning", "reinforcement-learning", "model-index", "region:us" ]
reinforcement-learning
jmurphy97
null
null
jmurphy97/dqn-SpaceInvadersNoFrameskip-v4
0
2
stable-baselines3
2023-04-10T20:51:34
--- library_name: stable-baselines3 tags: - SpaceInvadersNoFrameskip-v4 - deep-reinforcement-learning - reinforcement-learning - stable-baselines3 model-index: - name: DQN results: - task: type: reinforcement-learning name: reinforcement-learning dataset: name: SpaceInvadersNoFrameskip-v4 type: SpaceInvadersNoFrameskip-v4 metrics: - type: mean_reward value: 659.00 +/- 313.02 name: mean_reward verified: false --- # **DQN** Agent playing **SpaceInvadersNoFrameskip-v4** This is a trained model of a **DQN** agent playing **SpaceInvadersNoFrameskip-v4** using the [stable-baselines3 library](https://github.com/DLR-RM/stable-baselines3) and the [RL Zoo](https://github.com/DLR-RM/rl-baselines3-zoo). The RL Zoo is a training framework for Stable Baselines3 reinforcement learning agents, with hyperparameter optimization and pre-trained agents included. ## Usage (with SB3 RL Zoo) RL Zoo: https://github.com/DLR-RM/rl-baselines3-zoo<br/> SB3: https://github.com/DLR-RM/stable-baselines3<br/> SB3 Contrib: https://github.com/Stable-Baselines-Team/stable-baselines3-contrib Install the RL Zoo (with SB3 and SB3-Contrib): ```bash pip install rl_zoo3 ``` ``` # Download model and save it into the logs/ folder python -m rl_zoo3.load_from_hub --algo dqn --env SpaceInvadersNoFrameskip-v4 -orga jmurphy97 -f logs/ python -m rl_zoo3.enjoy --algo dqn --env SpaceInvadersNoFrameskip-v4 -f logs/ ``` If you installed the RL Zoo3 via pip (`pip install rl_zoo3`), from anywhere you can do: ``` python -m rl_zoo3.load_from_hub --algo dqn --env SpaceInvadersNoFrameskip-v4 -orga jmurphy97 -f logs/ python -m rl_zoo3.enjoy --algo dqn --env SpaceInvadersNoFrameskip-v4 -f logs/ ``` ## Training (with the RL Zoo) ``` python -m rl_zoo3.train --algo dqn --env SpaceInvadersNoFrameskip-v4 -f logs/ # Upload the model and generate video (when possible) python -m rl_zoo3.push_to_hub --algo dqn --env SpaceInvadersNoFrameskip-v4 -f logs/ -orga jmurphy97 ``` ## Hyperparameters ```python OrderedDict([('batch_size', 32), ('buffer_size', 100000), ('env_wrapper', ['stable_baselines3.common.atari_wrappers.AtariWrapper']), ('exploration_final_eps', 0.01), ('exploration_fraction', 0.1), ('frame_stack', 4), ('gradient_steps', 1), ('learning_rate', 0.0001), ('learning_starts', 100000), ('n_timesteps', 1000000.0), ('optimize_memory_usage', False), ('policy', 'CnnPolicy'), ('target_update_interval', 1000), ('train_freq', 4), ('normalize', False)]) ```
2,694
[ [ -0.04095458984375, -0.035400390625, 0.0214691162109375, 0.0246734619140625, -0.0094451904296875, -0.0165557861328125, 0.0128021240234375, -0.0142364501953125, 0.01317596435546875, 0.0242156982421875, -0.07135009765625, -0.035003662109375, -0.027435302734375, ...
jprorama/distilbert-base-uncased-finetuned-emotion
2023-04-11T10:47:16.000Z
[ "transformers", "pytorch", "distilbert", "text-classification", "generated_from_trainer", "dataset:emotion", "license:apache-2.0", "model-index", "endpoints_compatible", "region:us" ]
text-classification
jprorama
null
null
jprorama/distilbert-base-uncased-finetuned-emotion
0
2
transformers
2023-04-10T21:58:48
--- 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 config: split split: train args: split metrics: - name: Accuracy type: accuracy value: 0.9255 - name: F1 type: f1 value: 0.9254084497083122 --- <!-- 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: 0.2237 - Accuracy: 0.9255 - F1: 0.9254 ## 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 | |:-------------:|:-----:|:----:|:---------------:|:--------:|:------:| | 0.8639 | 1.0 | 250 | 0.3347 | 0.902 | 0.8993 | | 0.2552 | 2.0 | 500 | 0.2237 | 0.9255 | 0.9254 | ### Framework versions - Transformers 4.24.0 - Pytorch 2.0.0 - Datasets 2.10.1 - Tokenizers 0.11.0
1,837
[ [ -0.03765869140625, -0.04071044921875, 0.014190673828125, 0.022674560546875, -0.0260162353515625, -0.02001953125, -0.01273345947265625, -0.0089874267578125, 0.0098876953125, 0.008575439453125, -0.056365966796875, -0.051605224609375, -0.059417724609375, -0.007...
zhangzeyu/CT-PubMedBERT-RE-fine-tuned-noentity
2023-04-20T10:05:59.000Z
[ "transformers", "pytorch", "bert", "text-classification", "license:mit", "region:us" ]
text-classification
zhangzeyu
null
null
zhangzeyu/CT-PubMedBERT-RE-fine-tuned-noentity
1
2
transformers
2023-04-11T02:10:29
--- license: mit inference: false --- | Code | Ralation name | |------|----------------------------------------------------| | 0 | not_a_relation | | 1 | active_metabolites_of | | 2 | anatomic_structure_has_location | | 3 | anatomic_structure_is_physical_part_of | | 4 | anatomy_originated_from_biological_process | | 5 | associated_with_malfunction_of_gene_product | | 6 | biological_process_has_associated_location | | 7 | biological_process_has_initiator_chemical_or_drug | | 8 | biological_process_has_initiator_process | | 9 | biological_process_has_result_anatomy | | 10 | biological_process_has_result_biological_process | | 11 | biological_process_has_result_chemical_or_drug | | 12 | biological_process_involves_gene_product | | 13 | biological_process_is_part_of_process | | 14 | biological_process_results_from_biological_process | | 15 | biomarker_type_includes_gene_product | | 16 | cdrh_parent_of | | 17 | chemical_or_drug_affects_gene_product | | 18 | chemical_or_drug_initiates_biological_process | | 19 | chemical_or_drug_is_product_of_biological_process | | 20 | chemical_structure_of | | 21 | chemotherapy_regimen_has_component | | 22 | completely_excised_anatomy_has_procedure | | 23 | complex_has_physical_part | | 24 | concept_in_subset | | 25 | conceptual_part_of | | 26 | contraindicated_with_disease | | 27 | contraindicating_class_of | | 28 | disease_excludes_normal_cell_origin | | 29 | disease_excludes_primary_anatomic_site | | 30 | disease_has_abnormal_cell | | 31 | disease_has_associated_anatomic_site | | 32 | disease_has_associated_disease | | 33 | disease_has_associated_gene | | 34 | disease_has_finding | | 35 | disease_has_metastatic_anatomic_site | | 36 | disease_has_normal_cell_origin | | 37 | disease_has_normal_tissue_origin | | 38 | disease_has_primary_anatomic_site | | 39 | disease_may_have_associated_disease | | 40 | disease_may_have_finding | | 41 | excised_anatomy_has_procedure | | 42 | gene_associated_with_disease | | 43 | gene_encodes_gene_product | | 44 | gene_found_in_organism | | 45 | gene_mapped_to_disease | | 46 | gene_plays_role_in_process | | 47 | gene_product_affected_by_chemical_or_drug | | 48 | gene_product_encoded_by_gene | | 49 | gene_product_expressed_in_tissue | | 50 | gene_product_has_associated_anatomy | | 51 | gene_product_has_biochemical_function | | 52 | gene_product_has_chemical_classification | | 53 | gene_product_has_organism_source | | 54 | gene_product_has_structural_domain_or_motif | | 55 | gene_product_is_biomarker_of | | 56 | gene_product_is_physical_part_of | | 57 | gene_product_malfunction_associated_with_disease | | 58 | gene_product_plays_role_in_biological_process | | 59 | has_active_metabolites | | 60 | has_cdrh_parent | | 61 | has_chemical_structure | | 62 | has_conceptual_part | | 63 | has_contraindicated_drug | | 64 | has_contraindicating_class | | 65 | has_free_acid_or_base_form | | 66 | has_ingredient | | 67 | has_mechanism_of_action | | 68 | has_nichd_parent | | 69 | has_physical_part_of_anatomic_structure | | 70 | has_physiologic_effect | | 71 | has_salt_form | | 72 | has_therapeutic_class | | 73 | has_tradename | | 74 | induced_by | | 75 | induces | | 76 | ingredient_of | | 77 | is_abnormal_cell_of_disease | | 78 | is_associated_anatomic_site_of | | 79 | is_associated_anatomy_of_gene_product | | 80 | is_associated_disease_of | | 81 | is_biochemical_function_of_gene_product | | 82 | is_chemical_classification_of_gene_product | | 83 | is_component_of_chemotherapy_regimen | | 84 | is_finding_of_disease | | 85 | is_location_of_anatomic_structure | | 86 | is_location_of_biological_process | | 87 | is_marked_by_gene_product | | 88 | is_metastatic_anatomic_site_of_disease | | 89 | is_normal_cell_origin_of_disease | | 90 | is_normal_tissue_origin_of_disease | | 91 | is_not_normal_cell_origin_of_disease | | 92 | is_not_primary_anatomic_site_of_disease | | 93 | is_organism_source_of_gene_product | | 94 | is_physiologic_effect_of_chemical_or_drug | | 95 | is_primary_anatomic_site_of_disease | | 96 | is_structural_domain_or_motif_of_gene_product | | 97 | may_be_associated_disease_of_disease | | 98 | may_be_diagnosed_by | | 99 | may_be_finding_of_disease | | 100 | may_be_prevented_by | | 101 | may_be_treated_by | | 102 | may_diagnose | | 103 | may_prevent | | 104 | may_treat | | 105 | mechanism_of_action_of | | 106 | nichd_parent_of | | 107 | organism_has_gene | | 108 | partially_excised_anatomy_has_procedure | | 109 | pathogenesis_of_disease_involves_gene | | 110 | physiologic_effect_of | | 111 | procedure_has_completely_excised_anatomy | | 112 | procedure_has_excised_anatomy | | 113 | procedure_has_partially_excised_anatomy | | 114 | procedure_has_target_anatomy | | 115 | process_includes_biological_process | | 116 | process_initiates_biological_process | | 117 | process_involves_gene | | 118 | product_component_of | | 119 | special_category_includes_neoplasm | | 120 | subset_includes_concept | | 121 | target_anatomy_has_procedure | | 122 | therapeutic_class_of | | 123 | tissue_is_expression_site_of_gene_product | | 124 | tradename_of |
7,912
[ [ -0.029052734375, -0.034332275390625, 0.01824951171875, 0.025115966796875, -0.00799560546875, 0.0270843505859375, 0.016815185546875, -0.0228118896484375, 0.0657958984375, 0.02484130859375, -0.043365478515625, -0.062408447265625, -0.05755615234375, 0.028457641...
zhangzeyu/CT-PubMedBERT-RE-fine-tuned-group
2023-04-20T10:05:43.000Z
[ "transformers", "pytorch", "bert", "text-classification", "license:mit", "region:us" ]
text-classification
zhangzeyu
null
null
zhangzeyu/CT-PubMedBERT-RE-fine-tuned-group
0
2
transformers
2023-04-11T02:24:18
--- license: mit inference: false --- | Code | Ralation name | |------|----------------------------------------------------| | 0 | not_a_relation | | 1 | active_metabolites_of | | 2 | anatomic_structure_has_location | | 3 | anatomic_structure_is_physical_part_of | | 4 | anatomy_originated_from_biological_process | | 5 | associated_with_malfunction_of_gene_product | | 6 | biological_process_has_associated_location | | 7 | biological_process_has_initiator_chemical_or_drug | | 8 | biological_process_has_initiator_process | | 9 | biological_process_has_result_anatomy | | 10 | biological_process_has_result_biological_process | | 11 | biological_process_has_result_chemical_or_drug | | 12 | biological_process_involves_gene_product | | 13 | biological_process_is_part_of_process | | 14 | biological_process_results_from_biological_process | | 15 | biomarker_type_includes_gene_product | | 16 | cdrh_parent_of | | 17 | chemical_or_drug_affects_gene_product | | 18 | chemical_or_drug_initiates_biological_process | | 19 | chemical_or_drug_is_product_of_biological_process | | 20 | chemical_structure_of | | 21 | chemotherapy_regimen_has_component | | 22 | completely_excised_anatomy_has_procedure | | 23 | complex_has_physical_part | | 24 | concept_in_subset | | 25 | conceptual_part_of | | 26 | contraindicated_with_disease | | 27 | contraindicating_class_of | | 28 | disease_excludes_normal_cell_origin | | 29 | disease_excludes_primary_anatomic_site | | 30 | disease_has_abnormal_cell | | 31 | disease_has_associated_anatomic_site | | 32 | disease_has_associated_disease | | 33 | disease_has_associated_gene | | 34 | disease_has_finding | | 35 | disease_has_metastatic_anatomic_site | | 36 | disease_has_normal_cell_origin | | 37 | disease_has_normal_tissue_origin | | 38 | disease_has_primary_anatomic_site | | 39 | disease_may_have_associated_disease | | 40 | disease_may_have_finding | | 41 | excised_anatomy_has_procedure | | 42 | gene_associated_with_disease | | 43 | gene_encodes_gene_product | | 44 | gene_found_in_organism | | 45 | gene_mapped_to_disease | | 46 | gene_plays_role_in_process | | 47 | gene_product_affected_by_chemical_or_drug | | 48 | gene_product_encoded_by_gene | | 49 | gene_product_expressed_in_tissue | | 50 | gene_product_has_associated_anatomy | | 51 | gene_product_has_biochemical_function | | 52 | gene_product_has_chemical_classification | | 53 | gene_product_has_organism_source | | 54 | gene_product_has_structural_domain_or_motif | | 55 | gene_product_is_biomarker_of | | 56 | gene_product_is_physical_part_of | | 57 | gene_product_malfunction_associated_with_disease | | 58 | gene_product_plays_role_in_biological_process | | 59 | has_active_metabolites | | 60 | has_cdrh_parent | | 61 | has_chemical_structure | | 62 | has_conceptual_part | | 63 | has_contraindicated_drug | | 64 | has_contraindicating_class | | 65 | has_free_acid_or_base_form | | 66 | has_ingredient | | 67 | has_mechanism_of_action | | 68 | has_nichd_parent | | 69 | has_physical_part_of_anatomic_structure | | 70 | has_physiologic_effect | | 71 | has_salt_form | | 72 | has_therapeutic_class | | 73 | has_tradename | | 74 | induced_by | | 75 | induces | | 76 | ingredient_of | | 77 | is_abnormal_cell_of_disease | | 78 | is_associated_anatomic_site_of | | 79 | is_associated_anatomy_of_gene_product | | 80 | is_associated_disease_of | | 81 | is_biochemical_function_of_gene_product | | 82 | is_chemical_classification_of_gene_product | | 83 | is_component_of_chemotherapy_regimen | | 84 | is_finding_of_disease | | 85 | is_location_of_anatomic_structure | | 86 | is_location_of_biological_process | | 87 | is_marked_by_gene_product | | 88 | is_metastatic_anatomic_site_of_disease | | 89 | is_normal_cell_origin_of_disease | | 90 | is_normal_tissue_origin_of_disease | | 91 | is_not_normal_cell_origin_of_disease | | 92 | is_not_primary_anatomic_site_of_disease | | 93 | is_organism_source_of_gene_product | | 94 | is_physiologic_effect_of_chemical_or_drug | | 95 | is_primary_anatomic_site_of_disease | | 96 | is_structural_domain_or_motif_of_gene_product | | 97 | may_be_associated_disease_of_disease | | 98 | may_be_diagnosed_by | | 99 | may_be_finding_of_disease | | 100 | may_be_prevented_by | | 101 | may_be_treated_by | | 102 | may_diagnose | | 103 | may_prevent | | 104 | may_treat | | 105 | mechanism_of_action_of | | 106 | nichd_parent_of | | 107 | organism_has_gene | | 108 | partially_excised_anatomy_has_procedure | | 109 | pathogenesis_of_disease_involves_gene | | 110 | physiologic_effect_of | | 111 | procedure_has_completely_excised_anatomy | | 112 | procedure_has_excised_anatomy | | 113 | procedure_has_partially_excised_anatomy | | 114 | procedure_has_target_anatomy | | 115 | process_includes_biological_process | | 116 | process_initiates_biological_process | | 117 | process_involves_gene | | 118 | product_component_of | | 119 | special_category_includes_neoplasm | | 120 | subset_includes_concept | | 121 | target_anatomy_has_procedure | | 122 | therapeutic_class_of | | 123 | tissue_is_expression_site_of_gene_product | | 124 | tradename_of |
7,912
[ [ -0.0290374755859375, -0.034332275390625, 0.01824951171875, 0.025115966796875, -0.0079803466796875, 0.0270843505859375, 0.016815185546875, -0.0228271484375, 0.0657958984375, 0.02484130859375, -0.043365478515625, -0.062408447265625, -0.0574951171875, 0.0284423...
zhangzeyu/CT-PubMedBERT-RE-fine-tuned-type
2023-04-20T10:04:41.000Z
[ "transformers", "pytorch", "bert", "text-classification", "license:mit", "region:us" ]
text-classification
zhangzeyu
null
null
zhangzeyu/CT-PubMedBERT-RE-fine-tuned-type
0
2
transformers
2023-04-11T02:28:46
--- license: mit inference: false --- | Code | Ralation name | |------|----------------------------------------------------| | 0 | not_a_relation | | 1 | active_metabolites_of | | 2 | anatomic_structure_has_location | | 3 | anatomic_structure_is_physical_part_of | | 4 | anatomy_originated_from_biological_process | | 5 | associated_with_malfunction_of_gene_product | | 6 | biological_process_has_associated_location | | 7 | biological_process_has_initiator_chemical_or_drug | | 8 | biological_process_has_initiator_process | | 9 | biological_process_has_result_anatomy | | 10 | biological_process_has_result_biological_process | | 11 | biological_process_has_result_chemical_or_drug | | 12 | biological_process_involves_gene_product | | 13 | biological_process_is_part_of_process | | 14 | biological_process_results_from_biological_process | | 15 | biomarker_type_includes_gene_product | | 16 | cdrh_parent_of | | 17 | chemical_or_drug_affects_gene_product | | 18 | chemical_or_drug_initiates_biological_process | | 19 | chemical_or_drug_is_product_of_biological_process | | 20 | chemical_structure_of | | 21 | chemotherapy_regimen_has_component | | 22 | completely_excised_anatomy_has_procedure | | 23 | complex_has_physical_part | | 24 | concept_in_subset | | 25 | conceptual_part_of | | 26 | contraindicated_with_disease | | 27 | contraindicating_class_of | | 28 | disease_excludes_normal_cell_origin | | 29 | disease_excludes_primary_anatomic_site | | 30 | disease_has_abnormal_cell | | 31 | disease_has_associated_anatomic_site | | 32 | disease_has_associated_disease | | 33 | disease_has_associated_gene | | 34 | disease_has_finding | | 35 | disease_has_metastatic_anatomic_site | | 36 | disease_has_normal_cell_origin | | 37 | disease_has_normal_tissue_origin | | 38 | disease_has_primary_anatomic_site | | 39 | disease_may_have_associated_disease | | 40 | disease_may_have_finding | | 41 | excised_anatomy_has_procedure | | 42 | gene_associated_with_disease | | 43 | gene_encodes_gene_product | | 44 | gene_found_in_organism | | 45 | gene_mapped_to_disease | | 46 | gene_plays_role_in_process | | 47 | gene_product_affected_by_chemical_or_drug | | 48 | gene_product_encoded_by_gene | | 49 | gene_product_expressed_in_tissue | | 50 | gene_product_has_associated_anatomy | | 51 | gene_product_has_biochemical_function | | 52 | gene_product_has_chemical_classification | | 53 | gene_product_has_organism_source | | 54 | gene_product_has_structural_domain_or_motif | | 55 | gene_product_is_biomarker_of | | 56 | gene_product_is_physical_part_of | | 57 | gene_product_malfunction_associated_with_disease | | 58 | gene_product_plays_role_in_biological_process | | 59 | has_active_metabolites | | 60 | has_cdrh_parent | | 61 | has_chemical_structure | | 62 | has_conceptual_part | | 63 | has_contraindicated_drug | | 64 | has_contraindicating_class | | 65 | has_free_acid_or_base_form | | 66 | has_ingredient | | 67 | has_mechanism_of_action | | 68 | has_nichd_parent | | 69 | has_physical_part_of_anatomic_structure | | 70 | has_physiologic_effect | | 71 | has_salt_form | | 72 | has_therapeutic_class | | 73 | has_tradename | | 74 | induced_by | | 75 | induces | | 76 | ingredient_of | | 77 | is_abnormal_cell_of_disease | | 78 | is_associated_anatomic_site_of | | 79 | is_associated_anatomy_of_gene_product | | 80 | is_associated_disease_of | | 81 | is_biochemical_function_of_gene_product | | 82 | is_chemical_classification_of_gene_product | | 83 | is_component_of_chemotherapy_regimen | | 84 | is_finding_of_disease | | 85 | is_location_of_anatomic_structure | | 86 | is_location_of_biological_process | | 87 | is_marked_by_gene_product | | 88 | is_metastatic_anatomic_site_of_disease | | 89 | is_normal_cell_origin_of_disease | | 90 | is_normal_tissue_origin_of_disease | | 91 | is_not_normal_cell_origin_of_disease | | 92 | is_not_primary_anatomic_site_of_disease | | 93 | is_organism_source_of_gene_product | | 94 | is_physiologic_effect_of_chemical_or_drug | | 95 | is_primary_anatomic_site_of_disease | | 96 | is_structural_domain_or_motif_of_gene_product | | 97 | may_be_associated_disease_of_disease | | 98 | may_be_diagnosed_by | | 99 | may_be_finding_of_disease | | 100 | may_be_prevented_by | | 101 | may_be_treated_by | | 102 | may_diagnose | | 103 | may_prevent | | 104 | may_treat | | 105 | mechanism_of_action_of | | 106 | nichd_parent_of | | 107 | organism_has_gene | | 108 | partially_excised_anatomy_has_procedure | | 109 | pathogenesis_of_disease_involves_gene | | 110 | physiologic_effect_of | | 111 | procedure_has_completely_excised_anatomy | | 112 | procedure_has_excised_anatomy | | 113 | procedure_has_partially_excised_anatomy | | 114 | procedure_has_target_anatomy | | 115 | process_includes_biological_process | | 116 | process_initiates_biological_process | | 117 | process_involves_gene | | 118 | product_component_of | | 119 | special_category_includes_neoplasm | | 120 | subset_includes_concept | | 121 | target_anatomy_has_procedure | | 122 | therapeutic_class_of | | 123 | tissue_is_expression_site_of_gene_product | | 124 | tradename_of |
7,912
[ [ -0.0290374755859375, -0.034332275390625, 0.0182647705078125, 0.02508544921875, -0.00800323486328125, 0.0270843505859375, 0.016815185546875, -0.0228271484375, 0.0657958984375, 0.02484130859375, -0.043365478515625, -0.062408447265625, -0.05755615234375, 0.0284...
Muhsabrys/autotrain-mynaguib-48414117632
2023-04-11T02:44:27.000Z
[ "transformers", "pytorch", "bert", "text-classification", "autotrain", "ar", "dataset:Muhsabrys/autotrain-data-mynaguib", "co2_eq_emissions", "endpoints_compatible", "region:us" ]
text-classification
Muhsabrys
null
null
Muhsabrys/autotrain-mynaguib-48414117632
0
2
transformers
2023-04-11T02:43:16
--- tags: - autotrain - text-classification language: - ar widget: - text: "I love AutoTrain 🤗" datasets: - Muhsabrys/autotrain-data-mynaguib co2_eq_emissions: emissions: 0.510452418180777 --- # Model Trained Using AutoTrain - Problem type: Binary Classification - Model ID: 48414117632 - CO2 Emissions (in grams): 0.5105 ## Validation Metrics - Loss: 0.004 - Accuracy: 1.000 - Precision: 1.000 - Recall: 1.000 - AUC: 1.000 - F1: 1.000 ## 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/Muhsabrys/autotrain-mynaguib-48414117632 ``` Or Python API: ``` from transformers import AutoModelForSequenceClassification, AutoTokenizer model = AutoModelForSequenceClassification.from_pretrained("Muhsabrys/autotrain-mynaguib-48414117632", use_auth_token=True) tokenizer = AutoTokenizer.from_pretrained("Muhsabrys/autotrain-mynaguib-48414117632", use_auth_token=True) inputs = tokenizer("I love AutoTrain", return_tensors="pt") outputs = model(**inputs) ```
1,133
[ [ -0.031982421875, -0.025360107421875, 0.0148773193359375, 0.01419830322265625, 0.0007548332214355469, 0.0003619194030761719, 0.008880615234375, -0.0111236572265625, 0.0080718994140625, 0.01326751708984375, -0.062469482421875, -0.0328369140625, -0.05645751953125, ...
zhangzeyu/CT-PubMedBERT-RE-fine-tuned-groupabb
2023-04-20T10:05:21.000Z
[ "transformers", "pytorch", "bert", "text-classification", "license:mit", "region:us" ]
text-classification
zhangzeyu
null
null
zhangzeyu/CT-PubMedBERT-RE-fine-tuned-groupabb
0
2
transformers
2023-04-11T02:55:26
--- license: mit inference: false --- | Code | Ralation name | |------|----------------------------------------------------| | 0 | not_a_relation | | 1 | active_metabolites_of | | 2 | anatomic_structure_has_location | | 3 | anatomic_structure_is_physical_part_of | | 4 | anatomy_originated_from_biological_process | | 5 | associated_with_malfunction_of_gene_product | | 6 | biological_process_has_associated_location | | 7 | biological_process_has_initiator_chemical_or_drug | | 8 | biological_process_has_initiator_process | | 9 | biological_process_has_result_anatomy | | 10 | biological_process_has_result_biological_process | | 11 | biological_process_has_result_chemical_or_drug | | 12 | biological_process_involves_gene_product | | 13 | biological_process_is_part_of_process | | 14 | biological_process_results_from_biological_process | | 15 | biomarker_type_includes_gene_product | | 16 | cdrh_parent_of | | 17 | chemical_or_drug_affects_gene_product | | 18 | chemical_or_drug_initiates_biological_process | | 19 | chemical_or_drug_is_product_of_biological_process | | 20 | chemical_structure_of | | 21 | chemotherapy_regimen_has_component | | 22 | completely_excised_anatomy_has_procedure | | 23 | complex_has_physical_part | | 24 | concept_in_subset | | 25 | conceptual_part_of | | 26 | contraindicated_with_disease | | 27 | contraindicating_class_of | | 28 | disease_excludes_normal_cell_origin | | 29 | disease_excludes_primary_anatomic_site | | 30 | disease_has_abnormal_cell | | 31 | disease_has_associated_anatomic_site | | 32 | disease_has_associated_disease | | 33 | disease_has_associated_gene | | 34 | disease_has_finding | | 35 | disease_has_metastatic_anatomic_site | | 36 | disease_has_normal_cell_origin | | 37 | disease_has_normal_tissue_origin | | 38 | disease_has_primary_anatomic_site | | 39 | disease_may_have_associated_disease | | 40 | disease_may_have_finding | | 41 | excised_anatomy_has_procedure | | 42 | gene_associated_with_disease | | 43 | gene_encodes_gene_product | | 44 | gene_found_in_organism | | 45 | gene_mapped_to_disease | | 46 | gene_plays_role_in_process | | 47 | gene_product_affected_by_chemical_or_drug | | 48 | gene_product_encoded_by_gene | | 49 | gene_product_expressed_in_tissue | | 50 | gene_product_has_associated_anatomy | | 51 | gene_product_has_biochemical_function | | 52 | gene_product_has_chemical_classification | | 53 | gene_product_has_organism_source | | 54 | gene_product_has_structural_domain_or_motif | | 55 | gene_product_is_biomarker_of | | 56 | gene_product_is_physical_part_of | | 57 | gene_product_malfunction_associated_with_disease | | 58 | gene_product_plays_role_in_biological_process | | 59 | has_active_metabolites | | 60 | has_cdrh_parent | | 61 | has_chemical_structure | | 62 | has_conceptual_part | | 63 | has_contraindicated_drug | | 64 | has_contraindicating_class | | 65 | has_free_acid_or_base_form | | 66 | has_ingredient | | 67 | has_mechanism_of_action | | 68 | has_nichd_parent | | 69 | has_physical_part_of_anatomic_structure | | 70 | has_physiologic_effect | | 71 | has_salt_form | | 72 | has_therapeutic_class | | 73 | has_tradename | | 74 | induced_by | | 75 | induces | | 76 | ingredient_of | | 77 | is_abnormal_cell_of_disease | | 78 | is_associated_anatomic_site_of | | 79 | is_associated_anatomy_of_gene_product | | 80 | is_associated_disease_of | | 81 | is_biochemical_function_of_gene_product | | 82 | is_chemical_classification_of_gene_product | | 83 | is_component_of_chemotherapy_regimen | | 84 | is_finding_of_disease | | 85 | is_location_of_anatomic_structure | | 86 | is_location_of_biological_process | | 87 | is_marked_by_gene_product | | 88 | is_metastatic_anatomic_site_of_disease | | 89 | is_normal_cell_origin_of_disease | | 90 | is_normal_tissue_origin_of_disease | | 91 | is_not_normal_cell_origin_of_disease | | 92 | is_not_primary_anatomic_site_of_disease | | 93 | is_organism_source_of_gene_product | | 94 | is_physiologic_effect_of_chemical_or_drug | | 95 | is_primary_anatomic_site_of_disease | | 96 | is_structural_domain_or_motif_of_gene_product | | 97 | may_be_associated_disease_of_disease | | 98 | may_be_diagnosed_by | | 99 | may_be_finding_of_disease | | 100 | may_be_prevented_by | | 101 | may_be_treated_by | | 102 | may_diagnose | | 103 | may_prevent | | 104 | may_treat | | 105 | mechanism_of_action_of | | 106 | nichd_parent_of | | 107 | organism_has_gene | | 108 | partially_excised_anatomy_has_procedure | | 109 | pathogenesis_of_disease_involves_gene | | 110 | physiologic_effect_of | | 111 | procedure_has_completely_excised_anatomy | | 112 | procedure_has_excised_anatomy | | 113 | procedure_has_partially_excised_anatomy | | 114 | procedure_has_target_anatomy | | 115 | process_includes_biological_process | | 116 | process_initiates_biological_process | | 117 | process_involves_gene | | 118 | product_component_of | | 119 | special_category_includes_neoplasm | | 120 | subset_includes_concept | | 121 | target_anatomy_has_procedure | | 122 | therapeutic_class_of | | 123 | tissue_is_expression_site_of_gene_product | | 124 | tradename_of |
7,912
[ [ -0.0290374755859375, -0.034332275390625, 0.0182647705078125, 0.02508544921875, -0.00800323486328125, 0.0270843505859375, 0.016815185546875, -0.0228271484375, 0.0657958984375, 0.02484130859375, -0.043365478515625, -0.062408447265625, -0.05755615234375, 0.0284...
zhangzeyu/CT-PubMedBERT-RE-fine-tuned-typecode
2023-04-20T10:04:02.000Z
[ "transformers", "pytorch", "bert", "text-classification", "license:mit", "region:us" ]
text-classification
zhangzeyu
null
null
zhangzeyu/CT-PubMedBERT-RE-fine-tuned-typecode
0
2
transformers
2023-04-11T02:57:45
--- license: mit inference: false --- | Code | Ralation name | |------|----------------------------------------------------| | 0 | not_a_relation | | 1 | active_metabolites_of | | 2 | anatomic_structure_has_location | | 3 | anatomic_structure_is_physical_part_of | | 4 | anatomy_originated_from_biological_process | | 5 | associated_with_malfunction_of_gene_product | | 6 | biological_process_has_associated_location | | 7 | biological_process_has_initiator_chemical_or_drug | | 8 | biological_process_has_initiator_process | | 9 | biological_process_has_result_anatomy | | 10 | biological_process_has_result_biological_process | | 11 | biological_process_has_result_chemical_or_drug | | 12 | biological_process_involves_gene_product | | 13 | biological_process_is_part_of_process | | 14 | biological_process_results_from_biological_process | | 15 | biomarker_type_includes_gene_product | | 16 | cdrh_parent_of | | 17 | chemical_or_drug_affects_gene_product | | 18 | chemical_or_drug_initiates_biological_process | | 19 | chemical_or_drug_is_product_of_biological_process | | 20 | chemical_structure_of | | 21 | chemotherapy_regimen_has_component | | 22 | completely_excised_anatomy_has_procedure | | 23 | complex_has_physical_part | | 24 | concept_in_subset | | 25 | conceptual_part_of | | 26 | contraindicated_with_disease | | 27 | contraindicating_class_of | | 28 | disease_excludes_normal_cell_origin | | 29 | disease_excludes_primary_anatomic_site | | 30 | disease_has_abnormal_cell | | 31 | disease_has_associated_anatomic_site | | 32 | disease_has_associated_disease | | 33 | disease_has_associated_gene | | 34 | disease_has_finding | | 35 | disease_has_metastatic_anatomic_site | | 36 | disease_has_normal_cell_origin | | 37 | disease_has_normal_tissue_origin | | 38 | disease_has_primary_anatomic_site | | 39 | disease_may_have_associated_disease | | 40 | disease_may_have_finding | | 41 | excised_anatomy_has_procedure | | 42 | gene_associated_with_disease | | 43 | gene_encodes_gene_product | | 44 | gene_found_in_organism | | 45 | gene_mapped_to_disease | | 46 | gene_plays_role_in_process | | 47 | gene_product_affected_by_chemical_or_drug | | 48 | gene_product_encoded_by_gene | | 49 | gene_product_expressed_in_tissue | | 50 | gene_product_has_associated_anatomy | | 51 | gene_product_has_biochemical_function | | 52 | gene_product_has_chemical_classification | | 53 | gene_product_has_organism_source | | 54 | gene_product_has_structural_domain_or_motif | | 55 | gene_product_is_biomarker_of | | 56 | gene_product_is_physical_part_of | | 57 | gene_product_malfunction_associated_with_disease | | 58 | gene_product_plays_role_in_biological_process | | 59 | has_active_metabolites | | 60 | has_cdrh_parent | | 61 | has_chemical_structure | | 62 | has_conceptual_part | | 63 | has_contraindicated_drug | | 64 | has_contraindicating_class | | 65 | has_free_acid_or_base_form | | 66 | has_ingredient | | 67 | has_mechanism_of_action | | 68 | has_nichd_parent | | 69 | has_physical_part_of_anatomic_structure | | 70 | has_physiologic_effect | | 71 | has_salt_form | | 72 | has_therapeutic_class | | 73 | has_tradename | | 74 | induced_by | | 75 | induces | | 76 | ingredient_of | | 77 | is_abnormal_cell_of_disease | | 78 | is_associated_anatomic_site_of | | 79 | is_associated_anatomy_of_gene_product | | 80 | is_associated_disease_of | | 81 | is_biochemical_function_of_gene_product | | 82 | is_chemical_classification_of_gene_product | | 83 | is_component_of_chemotherapy_regimen | | 84 | is_finding_of_disease | | 85 | is_location_of_anatomic_structure | | 86 | is_location_of_biological_process | | 87 | is_marked_by_gene_product | | 88 | is_metastatic_anatomic_site_of_disease | | 89 | is_normal_cell_origin_of_disease | | 90 | is_normal_tissue_origin_of_disease | | 91 | is_not_normal_cell_origin_of_disease | | 92 | is_not_primary_anatomic_site_of_disease | | 93 | is_organism_source_of_gene_product | | 94 | is_physiologic_effect_of_chemical_or_drug | | 95 | is_primary_anatomic_site_of_disease | | 96 | is_structural_domain_or_motif_of_gene_product | | 97 | may_be_associated_disease_of_disease | | 98 | may_be_diagnosed_by | | 99 | may_be_finding_of_disease | | 100 | may_be_prevented_by | | 101 | may_be_treated_by | | 102 | may_diagnose | | 103 | may_prevent | | 104 | may_treat | | 105 | mechanism_of_action_of | | 106 | nichd_parent_of | | 107 | organism_has_gene | | 108 | partially_excised_anatomy_has_procedure | | 109 | pathogenesis_of_disease_involves_gene | | 110 | physiologic_effect_of | | 111 | procedure_has_completely_excised_anatomy | | 112 | procedure_has_excised_anatomy | | 113 | procedure_has_partially_excised_anatomy | | 114 | procedure_has_target_anatomy | | 115 | process_includes_biological_process | | 116 | process_initiates_biological_process | | 117 | process_involves_gene | | 118 | product_component_of | | 119 | special_category_includes_neoplasm | | 120 | subset_includes_concept | | 121 | target_anatomy_has_procedure | | 122 | therapeutic_class_of | | 123 | tissue_is_expression_site_of_gene_product | | 124 | tradename_of |
7,912
[ [ -0.0290374755859375, -0.034332275390625, 0.0182647705078125, 0.02508544921875, -0.00800323486328125, 0.0270843505859375, 0.016815185546875, -0.0228271484375, 0.0657958984375, 0.02484130859375, -0.043365478515625, -0.062408447265625, -0.05755615234375, 0.0284...
dingzhaohan/distilbert-base-uncased-finetuned-cola
2023-04-13T08:39:07.000Z
[ "transformers", "pytorch", "tensorboard", "distilbert", "text-classification", "generated_from_trainer", "dataset:glue", "license:apache-2.0", "model-index", "endpoints_compatible", "region:us" ]
text-classification
dingzhaohan
null
null
dingzhaohan/distilbert-base-uncased-finetuned-cola
0
2
transformers
2023-04-11T05:47:49
--- 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 config: cola split: validation args: cola metrics: - name: Matthews Correlation type: matthews_correlation value: 0.10315004767907714 --- <!-- 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: 0.6976 - Matthews Correlation: 0.1032 ## 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 | |:-------------:|:-----:|:----:|:---------------:|:--------------------:| | 0.6135 | 1.0 | 535 | 0.6257 | 0.0 | | 0.6078 | 2.0 | 1070 | 0.6187 | 0.0 | | 0.6038 | 3.0 | 1605 | 0.6179 | -0.0041 | | 0.5649 | 4.0 | 2140 | 0.6509 | 0.1006 | | 0.5093 | 5.0 | 2675 | 0.6976 | 0.1032 | ### Framework versions - Transformers 4.29.0.dev0 - Pytorch 1.13.1+cu116 - Datasets 2.10.1 - Tokenizers 0.13.2
2,049
[ [ -0.0236358642578125, -0.04949951171875, 0.00836944580078125, 0.0197601318359375, -0.0189361572265625, -0.00893402099609375, -0.005962371826171875, -0.00368499755859375, 0.0227508544921875, 0.01006317138671875, -0.0452880859375, -0.0374755859375, -0.0631713867187...
hoang14/pegasus-finetuned-samsum
2023-04-11T10:20:05.000Z
[ "transformers", "pytorch", "pegasus", "text2text-generation", "en", "dataset:samsum", "license:apache-2.0", "autotrain_compatible", "endpoints_compatible", "region:us" ]
text2text-generation
hoang14
null
null
hoang14/pegasus-finetuned-samsum
0
2
transformers
2023-04-11T09:39:45
--- license: apache-2.0 language: - en metrics: - rouge datasets: - samsum pipeline_tag: text2text-generation --- Summarization model based on pegasus, finetuned on samsum dataset source code: https://colab.research.google.com/drive/1FxdOV1fiHY3JC6dFw5T-NED1J8dKKHSO#scrollTo=pgdQ2up7vJoU metrics on samsum dataset: - rouge1: 0.436239 - rouge2: 0.209266 - rougeL: 0.34446 - rougeLsum: 0.344428
400
[ [ -0.021728515625, -0.03564453125, 0.02001953125, 0.02081298828125, -0.040924072265625, -0.03125, 0.026123046875, 0.0024967193603515625, 0.08050537109375, 0.059173583984375, -0.05419921875, -0.03662109375, -0.052642822265625, -0.01617431640625, -0.03527832...
cybersyn/robertuito-homomex-track1
2023-04-24T15:58:22.000Z
[ "transformers", "tf", "roberta", "text-classification", "generated_from_keras_callback", "endpoints_compatible", "region:us" ]
text-classification
cybersyn
null
null
cybersyn/robertuito-homomex-track1
0
2
transformers
2023-04-11T10:50:10
--- tags: - generated_from_keras_callback model-index: - name: robertuito-homomex-track1 results: [] --- <!-- This model card has been generated automatically according to the information Keras had access to. You should probably proofread and complete it, then remove this comment. --> # robertuito-homomex-track1 This model is a fine-tuned version of [pysentimiento/robertuito-base-uncased](https://huggingface.co/pysentimiento/robertuito-base-uncased) on an unknown dataset. It achieves the following results on the evaluation set: ## 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: - optimizer: {'name': 'AdamW', 'weight_decay': 0.004, 'clipnorm': None, 'global_clipnorm': None, 'clipvalue': None, 'use_ema': False, 'ema_momentum': 0.99, 'ema_overwrite_frequency': None, 'jit_compile': True, 'is_legacy_optimizer': False, 'learning_rate': {'class_name': 'PolynomialDecay', 'config': {'initial_learning_rate': 2e-05, 'decay_steps': 5600, 'end_learning_rate': 0.0, 'power': 1.0, 'cycle': False, 'name': None}}, 'beta_1': 0.9, 'beta_2': 0.999, 'epsilon': 1e-07, 'amsgrad': False} - training_precision: float32 ### Training results ### Framework versions - Transformers 4.28.1 - TensorFlow 2.12.0 - Datasets 2.11.0 - Tokenizers 0.13.3
1,467
[ [ -0.037811279296875, -0.037841796875, 0.0304107666015625, 0.012420654296875, -0.039276123046875, -0.009033203125, -0.018157958984375, -0.0201263427734375, 0.0218353271484375, 0.006805419921875, -0.0638427734375, -0.0543212890625, -0.059417724609375, -0.012481...
Augcos/ML-Agents-Pyramids
2023-04-11T10:50:30.000Z
[ "ml-agents", "tensorboard", "onnx", "Pyramids", "deep-reinforcement-learning", "reinforcement-learning", "ML-Agents-Pyramids", "region:us" ]
reinforcement-learning
Augcos
null
null
Augcos/ML-Agents-Pyramids
0
2
ml-agents
2023-04-11T10:50:25
--- library_name: ml-agents tags: - Pyramids - deep-reinforcement-learning - reinforcement-learning - ML-Agents-Pyramids --- # **ppo** Agent playing **Pyramids** This is a trained model of a **ppo** agent playing **Pyramids** using the [Unity ML-Agents Library](https://github.com/Unity-Technologies/ml-agents). ## Usage (with ML-Agents) The Documentation: https://github.com/huggingface/ml-agents#get-started We wrote a complete tutorial to learn to train your first agent using ML-Agents and publish it to the Hub: ### Resume the training ``` mlagents-learn <your_configuration_file_path.yaml> --run-id=<run_id> --resume ``` ### Watch your Agent play You can watch your agent **playing directly in your browser:**. 1. Go to https://huggingface.co/spaces/unity/ML-Agents-Pyramids 2. Step 1: Find your model_id: Augcos/ML-Agents-Pyramids 3. Step 2: Select your *.nn /*.onnx file 4. Click on Watch the agent play 👀
955
[ [ -0.0284576416015625, -0.02001953125, -0.0003368854522705078, 0.0272979736328125, -0.00910186767578125, 0.006649017333984375, 0.027984619140625, -0.00414276123046875, 0.035736083984375, 0.0362548828125, -0.036376953125, -0.051116943359375, -0.036041259765625, ...
szilard/bert-base-banking77-pt2
2023-04-11T14:32:20.000Z
[ "transformers", "pytorch", "tensorboard", "bert", "text-classification", "generated_from_trainer", "dataset:banking77", "license:apache-2.0", "model-index", "endpoints_compatible", "region:us" ]
text-classification
szilard
null
null
szilard/bert-base-banking77-pt2
0
2
transformers
2023-04-11T13:03:48
--- license: apache-2.0 tags: - generated_from_trainer datasets: - banking77 metrics: - f1 model-index: - name: bert-base-banking77-pt2 results: - task: name: Text Classification type: text-classification dataset: name: banking77 type: banking77 config: default split: test args: default metrics: - name: F1 type: f1 value: 0.9293371477596352 --- <!-- 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. --> # bert-base-banking77-pt2 This model is a fine-tuned version of [bert-base-uncased](https://huggingface.co/bert-base-uncased) on the banking77 dataset. It achieves the following results on the evaluation set: - Loss: 0.3046 - F1: 0.9293 ## 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: 16 - eval_batch_size: 8 - seed: 42 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 - lr_scheduler_type: linear - num_epochs: 3 ### Training results | Training Loss | Epoch | Step | Validation Loss | F1 | |:-------------:|:-----:|:----:|:---------------:|:------:| | 1.2034 | 1.0 | 626 | 0.8513 | 0.8310 | | 0.4223 | 2.0 | 1252 | 0.3760 | 0.9150 | | 0.2017 | 3.0 | 1878 | 0.3046 | 0.9293 | ### Framework versions - Transformers 4.27.1 - Pytorch 2.0.0+cu118 - Datasets 2.9.0 - Tokenizers 0.13.3
1,728
[ [ -0.0289764404296875, -0.03955078125, 0.010498046875, 0.013458251953125, -0.04443359375, -0.0262298583984375, -0.00881195068359375, -0.01837158203125, -0.0038127899169921875, 0.04083251953125, -0.0445556640625, -0.043548583984375, -0.052825927734375, -0.02839...
seanghay/whisper-small-khmer
2023-04-19T02:53:04.000Z
[ "transformers", "pytorch", "tensorboard", "onnx", "whisper", "automatic-speech-recognition", "hf-asr-leaderboard", "generated_from_trainer", "km", "dataset:openslr", "dataset:google/fleurs", "license:apache-2.0", "model-index", "endpoints_compatible", "region:us" ]
automatic-speech-recognition
seanghay
null
null
seanghay/whisper-small-khmer
1
2
transformers
2023-04-11T13:43:27
--- language: - km license: apache-2.0 tags: - hf-asr-leaderboard - generated_from_trainer datasets: - openslr - google/fleurs metrics: - wer model-index: - name: Whisper Small Khmer Spaced - Seanghay Yath results: - task: name: Automatic Speech Recognition type: automatic-speech-recognition dataset: name: Google FLEURS type: google/fleurs config: km_kh split: all metrics: - name: Wer type: wer value: 0.6464 --- <!-- 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. --> # whisper-small-khmer This model is a fine-tuned version of [openai/whisper-small](https://huggingface.co/openai/whisper-small) on the None dataset. It achieves the following results on the evaluation set: - Loss: 0.4657 - Wer: 0.6464 ## Model description This model is fine-tuned with Google FLEURS & OpenSLR (SLR42) dataset. - [ggml-model.bin](https://huggingface.co/seanghay/whisper-small-khmer/blob/main/ggml-model.bin) - [model.onnx](https://huggingface.co/seanghay/whisper-small-khmer/blob/main/model.onnx) ```python from transformers import pipeline pipe = pipeline( task="automatic-speech-recognition", model="seanghay/whisper-small-khmer", ) result = pipe("audio.wav", generate_kwargs={ "language":"<|km|>", "task":"transcribe"}, batch_size=16 ) print(result["text"]) ``` ## whisper.cpp ### 1. Transcode the input audio to 16kHz PCM ```shell ffmpeg -i audio.ogg -ar 16000 -ac 1 -c:a pcm_s16le output.wav ``` ### 2. Transcribe with whisper.cpp ```shell ./main -m ggml-model.bin -f output.wav --print-colors --language km ``` ## Training and evaluation data - `training` = google/fleurs['train+validation'] + openslr['train'] - `eval` = google/fleurs['test'] ## Training procedure This model was trained based on the project on [GitHub](https://github.com/seanghay/whisper-tiny-khmer) with an NVIDIA A10 24GB. ### Training hyperparameters The following hyperparameters were used during training: - learning_rate: 6.25e-06 - train_batch_size: 16 - eval_batch_size: 8 - seed: 42 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 - lr_scheduler_type: linear - lr_scheduler_warmup_steps: 800 - training_steps: 8000 - mixed_precision_training: Native AMP ### Training results | Training Loss | Epoch | Step | Validation Loss | Wer | |:-------------:|:-----:|:----:|:---------------:|:------:| | 0.2065 | 3.37 | 1000 | 0.3403 | 0.7929 | | 0.0446 | 6.73 | 2000 | 0.2911 | 0.6961 | | 0.008 | 10.1 | 3000 | 0.3578 | 0.6627 | | 0.003 | 13.47 | 4000 | 0.3982 | 0.6564 | | 0.0012 | 16.84 | 5000 | 0.4287 | 0.6512 | | 0.0004 | 20.2 | 6000 | 0.4499 | 0.6419 | | 0.0001 | 23.57 | 7000 | 0.4614 | 0.6469 | | 0.0001 | 26.94 | 8000 | 0.4657 | 0.6464 | ### Framework versions - Transformers 4.28.0.dev0 - Pytorch 2.0.0+cu117 - Datasets 2.11.1.dev0 - Tokenizers 0.13.3
3,122
[ [ -0.034698486328125, -0.046630859375, 0.0125579833984375, 0.0016651153564453125, -0.0210723876953125, -0.02142333984375, -0.0197601318359375, -0.031768798828125, 0.01035308837890625, 0.020111083984375, -0.04534912109375, -0.047210693359375, -0.053375244140625, ...
ku-accms/bert-base-japanese-ssuw
2023-04-12T04:40:42.000Z
[ "transformers", "pytorch", "bert", "fill-mask", "ja", "dataset:wikipedia", "license:cc-by-sa-4.0", "autotrain_compatible", "endpoints_compatible", "region:us" ]
fill-mask
ku-accms
null
null
ku-accms/bert-base-japanese-ssuw
1
2
transformers
2023-04-11T13:57:30
--- language: ja license: cc-by-sa-4.0 library_name: transformers tags: - bert - fill-mask datasets: - wikipedia mask_token: "[MASK]" widget: - text: "京都 大学 で [MASK] を 専攻 する 。" - text: "東京 は 日本 の [MASK] だ 。" - text: "カフェ で [MASK] を 注文 する 。" --- # ku-accms/bert-base-japanese-ssuw ## Model description This is a pre-trained Japanese BERT base model for super short unit words (SSUW). ## Pre-processing The input text should be converted to full-width (zenkaku) characters and segmented into super short unit words in advance (e.g., by KyTea). ## How to use You can use this model directly with a pipeline for masked language modeling: ```python >>> from transformers import pipeline >>> unmasker = pipeline('fill-mask', model='ku-accms/bert-base-japanese-ssuw') >>> unmasker("京都 大学 で [MASK] を 専攻 する 。") [{'sequence': '京都 大学 で 文学 を 専攻 する 。', 'score': '0.1464807540178299', 'token': '14603', 'token_str': '文学'} {'sequence': '京都 大学 で 哲学 を 専攻 する 。', 'score': '0.08064978569746017', 'token': '15917', 'token_str': '哲学'} {'sequence': '京都 大学 で 演劇 を 専攻 する 。', 'score': '0.0800977498292923', 'token': '16772', 'token_str': '演劇'} {'sequence': '京都 大学 で 法学 を 専攻 する 。', 'score': '0.04579947143793106', 'token': '16255', 'token_str': '法学'} {'sequence': '京都 大学 で 英語 を 専攻 する 。', 'score': '0.045536939054727554', 'token': '14592', 'token_str': '英語'} ``` Here is how to use this model to get the features of a given text in PyTorch: ```python import zenhan import Mykytea kytea_model_path = "somewhere" kytea = Mykytea.Mykytea("-model {} -notags".format(kytea_model_path)) def preprocess(text): return " ".join(kytea.getWS(zenhan.h2z(text))) from transformers import BertTokenizer, BertModel tokenizer = BertTokenizer.from_pretrained('ku-accms/bert-base-japanese-ssuw') model = BertModel.from_pretrained("ku-accms/bert-base-japanese-ssuw") text = "京都大学で自然言語処理を専攻する。" encoded_input = tokenizer(preprocess(text), return_tensors='pt') output = model(**encoded_input) ``` ## Training data We used a Japanese Wikipedia dump (as of 20230101, 3.3GB). ## Training procedure We first segmented the texts into words by KyTea and then tokenized the words into subwords using WordPiece with a vocabulary size of 32,000. We pre-trained the BERT model using [transformers](https://github.com/huggingface/transformers) library. The training took about 8 days using 4 NVIDIA A100-SXM4-80GB GPUs. The following hyperparameters were used for the pre-training. - learning_rate: 2e-4 - weight decay: 1e-2 - per_device_train_batch_size: 80 - num_devices: 4 - gradient_accumulation_steps: 3 - total_train_batch_size: 960 - max_seq_length: 512 - optimizer: AdamW with betas=(0.9,0.999) and epsilon=1e-06 - lr_scheduler_type: linear schedule with warmup - training_steps: 500,000 - warmup_steps: 10,000
2,819
[ [ -0.0291748046875, -0.06890869140625, 0.0295257568359375, 0.004558563232421875, -0.053192138671875, -0.007541656494140625, -0.03759765625, -0.019012451171875, 0.0268707275390625, 0.02728271484375, -0.044677734375, -0.041778564453125, -0.0421142578125, 0.00084...
amalik27/bert_ai
2023-04-11T21:05:45.000Z
[ "transformers", "pytorch", "tensorboard", "bert", "text-classification", "generated_from_trainer", "license:apache-2.0", "endpoints_compatible", "region:us" ]
text-classification
amalik27
null
null
amalik27/bert_ai
0
2
transformers
2023-04-11T14:58:11
--- license: apache-2.0 tags: - generated_from_trainer metrics: - accuracy - f1 - precision - recall model-index: - name: bert_ai 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. --> # bert_ai This model is a fine-tuned version of [bert-base-uncased](https://huggingface.co/bert-base-uncased) on the None dataset. It achieves the following results on the evaluation set: - Loss: 0.0761 - Accuracy: 0.9913 - F1: 0.9913 - Precision: 0.9833 - Recall: 0.9995 ## 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-08 - lr_scheduler_type: linear - num_epochs: 3 ### Training results | Training Loss | Epoch | Step | Validation Loss | Accuracy | F1 | Precision | Recall | |:-------------:|:-----:|:-----:|:---------------:|:--------:|:------:|:---------:|:------:| | 0.0358 | 1.0 | 6059 | 0.0390 | 0.9923 | 0.9923 | 0.9859 | 0.9989 | | 0.0187 | 2.0 | 12118 | 0.0738 | 0.9884 | 0.9884 | 0.9779 | 0.9993 | | 0.0056 | 3.0 | 18177 | 0.0761 | 0.9913 | 0.9913 | 0.9833 | 0.9995 | ### Framework versions - Transformers 4.27.4 - Pytorch 2.0.0+cu118 - Datasets 2.11.0 - Tokenizers 0.13.3
1,659
[ [ -0.034912109375, -0.044281005859375, 0.01508331298828125, 0.01024627685546875, -0.02252197265625, -0.031982421875, -0.01230621337890625, -0.0229339599609375, 0.01438140869140625, 0.0185394287109375, -0.053619384765625, -0.043243408203125, -0.04681396484375, ...
amalik27/bert_human
2023-04-12T00:30:10.000Z
[ "transformers", "pytorch", "tensorboard", "bert", "text-classification", "generated_from_trainer", "license:apache-2.0", "endpoints_compatible", "region:us" ]
text-classification
amalik27
null
null
amalik27/bert_human
0
2
transformers
2023-04-11T15:04:28
--- license: apache-2.0 tags: - generated_from_trainer metrics: - accuracy - f1 - precision - recall model-index: - name: bert_human 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. --> # bert_human This model is a fine-tuned version of [bert-base-uncased](https://huggingface.co/bert-base-uncased) on the None dataset. It achieves the following results on the evaluation set: - Loss: 0.0451 - Accuracy: 0.9930 - F1: 0.9930 - Precision: 0.9923 - Recall: 0.9921 ## 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-08 - lr_scheduler_type: linear - num_epochs: 3 ### Training results | Training Loss | Epoch | Step | Validation Loss | Accuracy | F1 | Precision | Recall | |:-------------:|:-----:|:-----:|:---------------:|:--------:|:------:|:---------:|:------:| | 0.062 | 1.0 | 5488 | 0.0409 | 0.9914 | 0.9914 | 0.9924 | 0.9885 | | 0.0279 | 2.0 | 10976 | 0.0414 | 0.9925 | 0.9925 | 0.9923 | 0.9909 | | 0.008 | 3.0 | 16464 | 0.0451 | 0.9930 | 0.9930 | 0.9923 | 0.9921 | ### Framework versions - Transformers 4.27.4 - Pytorch 2.0.0+cu118 - Datasets 2.11.0 - Tokenizers 0.13.3
1,665
[ [ -0.036376953125, -0.037445068359375, 0.01140594482421875, 0.0093841552734375, -0.0201416015625, -0.024200439453125, -0.01540374755859375, -0.024871826171875, 0.01495361328125, 0.0218048095703125, -0.053070068359375, -0.04541015625, -0.04241943359375, -0.0130...
abulatk1n/distilbert-base-uncased-finetuned-emotion
2023-04-11T18:26:31.000Z
[ "transformers", "pytorch", "tensorboard", "distilbert", "text-classification", "generated_from_trainer", "dataset:emotion", "license:apache-2.0", "model-index", "endpoints_compatible", "region:us" ]
text-classification
abulatk1n
null
null
abulatk1n/distilbert-base-uncased-finetuned-emotion
0
2
transformers
2023-04-11T18:07:02
--- 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 config: split split: validation args: split metrics: - name: Accuracy type: accuracy value: 0.9235 - name: F1 type: f1 value: 0.9233783185589441 --- <!-- 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: 0.2220 - Accuracy: 0.9235 - F1: 0.9234 ## 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 | |:-------------:|:-----:|:----:|:---------------:|:--------:|:------:| | 0.8018 | 1.0 | 250 | 0.3189 | 0.9025 | 0.8981 | | 0.2488 | 2.0 | 500 | 0.2220 | 0.9235 | 0.9234 | ### Framework versions - Transformers 4.27.4 - Pytorch 2.0.0+cu118 - Datasets 2.11.0 - Tokenizers 0.13.3
1,848
[ [ -0.03814697265625, -0.041351318359375, 0.0144805908203125, 0.022003173828125, -0.02557373046875, -0.01898193359375, -0.01306915283203125, -0.00859832763671875, 0.0105133056640625, 0.00833892822265625, -0.056640625, -0.05194091796875, -0.060302734375, -0.0086...
amalik27/bert_combo
2023-04-11T22:49:44.000Z
[ "transformers", "pytorch", "tensorboard", "bert", "text-classification", "generated_from_trainer", "license:apache-2.0", "endpoints_compatible", "region:us" ]
text-classification
amalik27
null
null
amalik27/bert_combo
0
2
transformers
2023-04-11T21:35:24
--- license: apache-2.0 tags: - generated_from_trainer metrics: - accuracy - f1 - precision - recall model-index: - name: bert_combo 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. --> # bert_combo This model is a fine-tuned version of [bert-base-uncased](https://huggingface.co/bert-base-uncased) on the None dataset. It achieves the following results on the evaluation set: - Loss: 0.0881 - Accuracy: 0.9862 - F1: 0.9862 - Precision: 0.9788 - Recall: 0.9940 ## 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-08 - lr_scheduler_type: linear - num_epochs: 3 ### Training results | Training Loss | Epoch | Step | Validation Loss | Accuracy | F1 | Precision | Recall | |:-------------:|:-----:|:-----:|:---------------:|:--------:|:------:|:---------:|:------:| | 0.0848 | 1.0 | 6059 | 0.0705 | 0.9834 | 0.9834 | 0.9903 | 0.9766 | | 0.0363 | 2.0 | 12118 | 0.0925 | 0.9821 | 0.9821 | 0.9701 | 0.9950 | | 0.0118 | 3.0 | 18177 | 0.0881 | 0.9862 | 0.9862 | 0.9788 | 0.9940 | ### Framework versions - Transformers 4.27.4 - Pytorch 2.0.0+cu118 - Datasets 2.11.0 - Tokenizers 0.13.3
1,665
[ [ -0.038360595703125, -0.032806396484375, 0.0114593505859375, 0.00930023193359375, -0.027923583984375, -0.0213165283203125, -0.0106201171875, -0.018768310546875, 0.02252197265625, 0.0252685546875, -0.051483154296875, -0.03875732421875, -0.047271728515625, -0.0...
ValenHumano/roberta-base-bne-detector-de-stress-detector-de-stress
2023-04-11T21:48:01.000Z
[ "transformers", "pytorch", "tensorboard", "roberta", "text-classification", "generated_from_trainer", "license:apache-2.0", "endpoints_compatible", "region:us" ]
text-classification
ValenHumano
null
null
ValenHumano/roberta-base-bne-detector-de-stress-detector-de-stress
0
2
transformers
2023-04-11T21:36:49
--- license: apache-2.0 tags: - generated_from_trainer metrics: - accuracy model-index: - name: roberta-base-bne-detector-de-stress-detector-de-stress 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. --> # roberta-base-bne-detector-de-stress-detector-de-stress This model is a fine-tuned version of [ValenHumano/roberta-base-bne-detector-de-stress](https://huggingface.co/ValenHumano/roberta-base-bne-detector-de-stress) on the None dataset. It achieves the following results on the evaluation set: - Loss: 0.4838 - Accuracy: 0.7571 ## 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 | |:-------------:|:-----:|:----:|:---------------:|:--------:| | 0.4735 | 1.0 | 169 | 0.3888 | 0.8143 | | 0.2484 | 2.0 | 338 | 0.4838 | 0.7571 | ### Framework versions - Transformers 4.27.4 - Pytorch 2.0.0+cu118 - Datasets 2.11.0 - Tokenizers 0.13.3
1,522
[ [ -0.03997802734375, -0.04522705078125, 0.0185546875, 0.00441741943359375, -0.031951904296875, -0.044708251953125, -0.005031585693359375, -0.027618408203125, 0.003734588623046875, 0.0198822021484375, -0.04638671875, -0.050811767578125, -0.056915283203125, -0.0...
rlucasz93/ppo-Pyramid
2023-04-11T22:15:13.000Z
[ "ml-agents", "tensorboard", "onnx", "Pyramids", "deep-reinforcement-learning", "reinforcement-learning", "ML-Agents-Pyramids", "region:us" ]
reinforcement-learning
rlucasz93
null
null
rlucasz93/ppo-Pyramid
0
2
ml-agents
2023-04-11T22:04:38
--- library_name: ml-agents tags: - Pyramids - deep-reinforcement-learning - reinforcement-learning - ML-Agents-Pyramids --- # **ppo** Agent playing **Pyramids** This is a trained model of a **ppo** agent playing **Pyramids** using the [Unity ML-Agents Library](https://github.com/Unity-Technologies/ml-agents). ## Usage (with ML-Agents) The Documentation: https://github.com/huggingface/ml-agents#get-started We wrote a complete tutorial to learn to train your first agent using ML-Agents and publish it to the Hub: ### Resume the training ``` mlagents-learn <your_configuration_file_path.yaml> --run-id=<run_id> --resume ``` ### Watch your Agent play You can watch your agent **playing directly in your browser:**. 1. Go to https://huggingface.co/spaces/unity/ML-Agents-Pyramids 2. Step 1: Find your model_id: rlucasz93/ppo-Pyramid 3. Step 2: Select your *.nn /*.onnx file 4. Click on Watch the agent play 👀
951
[ [ -0.02685546875, -0.0195770263671875, 0.00006020069122314453, 0.0263519287109375, -0.0104217529296875, 0.005649566650390625, 0.027984619140625, -0.0038394927978515625, 0.0347900390625, 0.03631591796875, -0.03607177734375, -0.051971435546875, -0.03631591796875, ...
feng5520/dqn-SpaceInvadersNoFrameskip-v4
2023-04-12T00:58:03.000Z
[ "stable-baselines3", "SpaceInvadersNoFrameskip-v4", "deep-reinforcement-learning", "reinforcement-learning", "model-index", "region:us" ]
reinforcement-learning
feng5520
null
null
feng5520/dqn-SpaceInvadersNoFrameskip-v4
0
2
stable-baselines3
2023-04-12T00:57:31
--- library_name: stable-baselines3 tags: - SpaceInvadersNoFrameskip-v4 - deep-reinforcement-learning - reinforcement-learning - stable-baselines3 model-index: - name: DQN results: - task: type: reinforcement-learning name: reinforcement-learning dataset: name: SpaceInvadersNoFrameskip-v4 type: SpaceInvadersNoFrameskip-v4 metrics: - type: mean_reward value: 15.50 +/- 12.54 name: mean_reward verified: false --- # **DQN** Agent playing **SpaceInvadersNoFrameskip-v4** This is a trained model of a **DQN** agent playing **SpaceInvadersNoFrameskip-v4** using the [stable-baselines3 library](https://github.com/DLR-RM/stable-baselines3) and the [RL Zoo](https://github.com/DLR-RM/rl-baselines3-zoo). The RL Zoo is a training framework for Stable Baselines3 reinforcement learning agents, with hyperparameter optimization and pre-trained agents included. ## Usage (with SB3 RL Zoo) RL Zoo: https://github.com/DLR-RM/rl-baselines3-zoo<br/> SB3: https://github.com/DLR-RM/stable-baselines3<br/> SB3 Contrib: https://github.com/Stable-Baselines-Team/stable-baselines3-contrib Install the RL Zoo (with SB3 and SB3-Contrib): ```bash pip install rl_zoo3 ``` ``` # Download model and save it into the logs/ folder python -m rl_zoo3.load_from_hub --algo dqn --env SpaceInvadersNoFrameskip-v4 -orga feng5520 -f logs/ python -m rl_zoo3.enjoy --algo dqn --env SpaceInvadersNoFrameskip-v4 -f logs/ ``` If you installed the RL Zoo3 via pip (`pip install rl_zoo3`), from anywhere you can do: ``` python -m rl_zoo3.load_from_hub --algo dqn --env SpaceInvadersNoFrameskip-v4 -orga feng5520 -f logs/ python -m rl_zoo3.enjoy --algo dqn --env SpaceInvadersNoFrameskip-v4 -f logs/ ``` ## Training (with the RL Zoo) ``` python -m rl_zoo3.train --algo dqn --env SpaceInvadersNoFrameskip-v4 -f logs/ # Upload the model and generate video (when possible) python -m rl_zoo3.push_to_hub --algo dqn --env SpaceInvadersNoFrameskip-v4 -f logs/ -orga feng5520 ``` ## Hyperparameters ```python OrderedDict([('batch_size', 32), ('buffer_size', 100000), ('env_wrapper', ['stable_baselines3.common.atari_wrappers.AtariWrapper']), ('exploration_final_eps', 0.01), ('exploration_fraction', 0.1), ('frame_stack', 4), ('gradient_steps', 1), ('learning_rate', 0.0001), ('learning_starts', 100000), ('n_timesteps', 10000.0), ('optimize_memory_usage', False), ('policy', 'CnnPolicy'), ('target_update_interval', 1000), ('train_freq', 4), ('normalize', False)]) ```
2,687
[ [ -0.04107666015625, -0.036712646484375, 0.022491455078125, 0.0255126953125, -0.0098724365234375, -0.01971435546875, 0.0117340087890625, -0.0134735107421875, 0.01285552978515625, 0.0250396728515625, -0.06927490234375, -0.03631591796875, -0.0272064208984375, -0...
willmendoza/platzi-distilroberta-base-mrpc-glue-will-mendoza
2023-04-12T01:30:43.000Z
[ "transformers", "pytorch", "tensorboard", "roberta", "text-classification", "generated_from_trainer", "dataset:glue", "license:apache-2.0", "model-index", "endpoints_compatible", "region:us" ]
text-classification
willmendoza
null
null
willmendoza/platzi-distilroberta-base-mrpc-glue-will-mendoza
0
2
transformers
2023-04-12T01:18:35
--- license: apache-2.0 tags: - text-classification - generated_from_trainer datasets: - glue metrics: - accuracy - f1 model-index: - name: platzi-distilroberta-base-mrpc-glue-will-mendoza results: - task: name: Text Classification type: text-classification dataset: name: datasetX type: glue config: mrpc split: validation args: mrpc metrics: - name: Accuracy type: accuracy value: 0.8382352941176471 - name: F1 type: f1 value: 0.8773234200743494 --- <!-- 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. --> # platzi-distilroberta-base-mrpc-glue-will-mendoza This model is a fine-tuned version of [distilroberta-base](https://huggingface.co/distilroberta-base) on the datasetX dataset. It achieves the following results on the evaluation set: - Loss: 0.5374 - Accuracy: 0.8382 - F1: 0.8773 ## 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: 3 ### Training results | Training Loss | Epoch | Step | Validation Loss | Accuracy | F1 | |:-------------:|:-----:|:----:|:---------------:|:--------:|:------:| | 0.5458 | 1.09 | 500 | 0.5644 | 0.8309 | 0.8832 | | 0.3627 | 2.18 | 1000 | 0.5374 | 0.8382 | 0.8773 | ### Framework versions - Transformers 4.27.4 - Pytorch 2.0.0+cu118 - Datasets 2.11.0 - Tokenizers 0.13.3
1,878
[ [ -0.0306243896484375, -0.037322998046875, 0.01297760009765625, 0.022186279296875, -0.02838134765625, -0.0249786376953125, -0.0085296630859375, -0.0024967193603515625, 0.00359344482421875, 0.016357421875, -0.049560546875, -0.046478271484375, -0.058837890625, -...
Nbardy/holycene-diffusers
2023-04-12T01:23:01.000Z
[ "diffusers", "region:us" ]
null
Nbardy
null
null
Nbardy/holycene-diffusers
0
2
diffusers
2023-04-12T01:23:06
https://civitai.com/models/24345 in diffusers format for compatibility All credits go to the original authors
111
[ [ -0.01058197021484375, 0.0034351348876953125, 0.057769775390625, 0.06256103515625, -0.01091766357421875, -0.025360107421875, 0.0253753662109375, 0.01218414306640625, 0.0131072998046875, 0.0311431884765625, -0.0173187255859375, 0.00672149658203125, -0.011199951171...
davidliu1110/bert-base-chinese-wikiann-zh-ner-2
2023-04-12T01:51:35.000Z
[ "transformers", "pytorch", "tensorboard", "bert", "token-classification", "generated_from_trainer", "dataset:wikiann", "model-index", "autotrain_compatible", "endpoints_compatible", "region:us" ]
token-classification
davidliu1110
null
null
davidliu1110/bert-base-chinese-wikiann-zh-ner-2
0
2
transformers
2023-04-12T01:32:18
--- tags: - generated_from_trainer datasets: - wikiann metrics: - precision - recall - f1 - accuracy model-index: - name: bert-base-chinese-wikiann-zh-ner-2 results: - task: name: Token Classification type: token-classification dataset: name: wikiann type: wikiann config: zh split: validation args: zh metrics: - name: Precision type: precision value: 0.7577054794520548 - name: Recall type: recall value: 0.7792363723685264 - name: F1 type: f1 value: 0.7683201136498164 - name: Accuracy type: accuracy value: 0.9385963268365817 --- <!-- 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. --> # bert-base-chinese-wikiann-zh-ner-2 This model is a fine-tuned version of [bert-base-chinese](https://huggingface.co/bert-base-chinese) on the wikiann dataset. It achieves the following results on the evaluation set: - Loss: 0.2036 - Precision: 0.7577 - Recall: 0.7792 - F1: 0.7683 - Accuracy: 0.9386 ## 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: 8 - eval_batch_size: 8 - seed: 42 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 - lr_scheduler_type: linear - lr_scheduler_warmup_ratio: 0.1 - num_epochs: 1 - mixed_precision_training: Native AMP ### Training results | Training Loss | Epoch | Step | Validation Loss | Precision | Recall | F1 | Accuracy | |:-------------:|:-----:|:----:|:---------------:|:---------:|:------:|:------:|:--------:| | 0.555 | 0.16 | 400 | 0.3120 | 0.5949 | 0.7117 | 0.6481 | 0.9041 | | 0.2944 | 0.32 | 800 | 0.2669 | 0.7013 | 0.7052 | 0.7032 | 0.9230 | | 0.2814 | 0.48 | 1200 | 0.2354 | 0.7078 | 0.7601 | 0.7330 | 0.9317 | | 0.2351 | 0.64 | 1600 | 0.2271 | 0.7295 | 0.7715 | 0.7499 | 0.9336 | | 0.2101 | 0.8 | 2000 | 0.2148 | 0.7478 | 0.7764 | 0.7618 | 0.9369 | | 0.23 | 0.96 | 2400 | 0.2059 | 0.7586 | 0.7752 | 0.7668 | 0.9385 | ### Framework versions - Transformers 4.27.4 - Pytorch 2.0.0+cu118 - Datasets 2.11.0 - Tokenizers 0.13.3
2,536
[ [ -0.033905029296875, -0.041107177734375, 0.0009784698486328125, 0.00940704345703125, -0.0213623046875, -0.0345458984375, -0.0136566162109375, -0.019866943359375, 0.0225677490234375, 0.021728515625, -0.049468994140625, -0.044586181640625, -0.044525146484375, -...
hlyu/nert_0dense
2023-04-12T01:42:17.000Z
[ "sentence-transformers", "pytorch", "bert", "feature-extraction", "sentence-similarity", "transformers", "endpoints_compatible", "region:us" ]
sentence-similarity
hlyu
null
null
hlyu/nert_0dense
0
2
sentence-transformers
2023-04-12T01:41:50
--- pipeline_tag: sentence-similarity tags: - sentence-transformers - feature-extraction - sentence-similarity - transformers --- # hlyu/nert_0dense This is a [sentence-transformers](https://www.SBERT.net) model: It maps sentences & paragraphs to a 768 dimensional dense vector space and can be used for tasks like clustering or semantic search. <!--- Describe your model here --> ## Usage (Sentence-Transformers) Using this model becomes easy when you have [sentence-transformers](https://www.SBERT.net) installed: ``` pip install -U sentence-transformers ``` Then you can use the model like this: ```python from sentence_transformers import SentenceTransformer sentences = ["This is an example sentence", "Each sentence is converted"] model = SentenceTransformer('hlyu/nert_0dense') embeddings = model.encode(sentences) print(embeddings) ``` ## Usage (HuggingFace Transformers) Without [sentence-transformers](https://www.SBERT.net), you can use the model like this: First, you pass your input through the transformer model, then you have to apply the right pooling-operation on-top of the contextualized word embeddings. ```python from transformers import AutoTokenizer, AutoModel import torch #Mean Pooling - Take attention mask into account for correct averaging def mean_pooling(model_output, attention_mask): token_embeddings = model_output[0] #First element of model_output contains all token embeddings input_mask_expanded = attention_mask.unsqueeze(-1).expand(token_embeddings.size()).float() return torch.sum(token_embeddings * input_mask_expanded, 1) / torch.clamp(input_mask_expanded.sum(1), min=1e-9) # Sentences we want sentence embeddings for sentences = ['This is an example sentence', 'Each sentence is converted'] # Load model from HuggingFace Hub tokenizer = AutoTokenizer.from_pretrained('hlyu/nert_0dense') model = AutoModel.from_pretrained('hlyu/nert_0dense') # Tokenize sentences encoded_input = tokenizer(sentences, padding=True, truncation=True, return_tensors='pt') # Compute token embeddings with torch.no_grad(): model_output = model(**encoded_input) # Perform pooling. In this case, mean pooling. sentence_embeddings = mean_pooling(model_output, encoded_input['attention_mask']) print("Sentence embeddings:") print(sentence_embeddings) ``` ## Evaluation Results <!--- Describe how your model was evaluated --> For an automated evaluation of this model, see the *Sentence Embeddings Benchmark*: [https://seb.sbert.net](https://seb.sbert.net?model_name=hlyu/nert_0dense) ## Training The model was trained with the parameters: **DataLoader**: `torch.utils.data.dataloader.DataLoader` of length 5055 with parameters: ``` {'batch_size': 128, 'sampler': 'torch.utils.data.sampler.RandomSampler', 'batch_sampler': 'torch.utils.data.sampler.BatchSampler'} ``` **Loss**: `sentence_transformers.losses.MSELoss.MSELoss` Parameters of the fit()-Method: ``` { "epochs": 1, "evaluation_steps": 1000, "evaluator": "sentence_transformers.evaluation.SequentialEvaluator.SequentialEvaluator", "max_grad_norm": 1, "optimizer_class": "<class 'torch.optim.adamw.AdamW'>", "optimizer_params": { "eps": 1e-06, "lr": 0.0001 }, "scheduler": "WarmupLinear", "steps_per_epoch": null, "warmup_steps": 1000, "weight_decay": 0.01 } ``` ## Full Model Architecture ``` SentenceTransformer( (0): Transformer({'max_seq_length': 512, 'do_lower_case': False}) with Transformer model: BertModel (1): Pooling({'word_embedding_dimension': 768, 'pooling_mode_cls_token': False, 'pooling_mode_mean_tokens': True, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False}) ) ``` ## Citing & Authors <!--- Describe where people can find more information -->
3,785
[ [ -0.0176544189453125, -0.06304931640625, 0.02069091796875, 0.0249481201171875, -0.015869140625, -0.032501220703125, -0.0181732177734375, 0.00017547607421875, 0.01922607421875, 0.0240020751953125, -0.049530029296875, -0.04278564453125, -0.050018310546875, 0.00...
tjayant/my_awesome_model_b
2023-04-12T05:08:08.000Z
[ "transformers", "tf", "bert", "text-classification", "generated_from_keras_callback", "license:apache-2.0", "endpoints_compatible", "region:us" ]
text-classification
tjayant
null
null
tjayant/my_awesome_model_b
0
2
transformers
2023-04-12T02:31:11
--- license: apache-2.0 tags: - generated_from_keras_callback model-index: - name: tjayant/my_awesome_model_b results: [] --- <!-- This model card has been generated automatically according to the information Keras had access to. You should probably proofread and complete it, then remove this comment. --> # tjayant/my_awesome_model_b This model is a fine-tuned version of [bert-base-uncased](https://huggingface.co/bert-base-uncased) on an unknown dataset. It achieves the following results on the evaluation set: - Train Loss: 1.2676 - Validation Loss: 1.2487 - Train Accuracy: 0.3923 - Epoch: 1 ## 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: - optimizer: {'name': 'Adam', 'weight_decay': None, 'clipnorm': None, 'global_clipnorm': None, 'clipvalue': None, 'use_ema': False, 'ema_momentum': 0.99, 'ema_overwrite_frequency': None, 'jit_compile': True, 'is_legacy_optimizer': False, 'learning_rate': {'class_name': 'PolynomialDecay', 'config': {'initial_learning_rate': 2e-05, 'decay_steps': 2280, 'end_learning_rate': 0.0, 'power': 1.0, 'cycle': False, 'name': None}}, 'beta_1': 0.9, 'beta_2': 0.999, 'epsilon': 1e-08, 'amsgrad': False} - training_precision: float32 ### Training results | Train Loss | Validation Loss | Train Accuracy | Epoch | |:----------:|:---------------:|:--------------:|:-----:| | 1.3131 | 1.2899 | 0.3832 | 0 | | 1.2676 | 1.2487 | 0.3923 | 1 | ### Framework versions - Transformers 4.27.4 - TensorFlow 2.12.0 - Datasets 2.11.0 - Tokenizers 0.13.3
1,761
[ [ -0.042236328125, -0.042999267578125, 0.0237884521484375, 0.007518768310546875, -0.0295562744140625, -0.0238037109375, -0.0137481689453125, -0.0242919921875, 0.01114654541015625, 0.004314422607421875, -0.04693603515625, -0.0491943359375, -0.051544189453125, -...
davidliu1110/bert-base-chinese-wikiann-zh-ner
2023-04-12T03:05:55.000Z
[ "transformers", "pytorch", "bert", "token-classification", "generated_from_trainer", "dataset:wikiann", "model-index", "autotrain_compatible", "endpoints_compatible", "region:us" ]
token-classification
davidliu1110
null
null
davidliu1110/bert-base-chinese-wikiann-zh-ner
0
2
transformers
2023-04-12T02:31:56
--- tags: - generated_from_trainer datasets: - wikiann metrics: - precision - recall - f1 - accuracy model-index: - name: bert-base-chinese-wikiann-zh-ner results: - task: name: Token Classification type: token-classification dataset: name: wikiann type: wikiann config: zh split: validation args: zh metrics: - name: Precision type: precision value: 0.7890612756621219 - name: Recall type: recall value: 0.8060513887777155 - name: F1 type: f1 value: 0.797465848346862 - name: Accuracy type: accuracy value: 0.9432393178410795 --- <!-- 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. --> # bert-base-chinese-wikiann-zh-ner This model is a fine-tuned version of [bert-base-chinese](https://huggingface.co/bert-base-chinese) on the wikiann dataset. It achieves the following results on the evaluation set: - Loss: 0.2092 - Precision: 0.7891 - Recall: 0.8061 - F1: 0.7975 - Accuracy: 0.9432 ## 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: 8 - eval_batch_size: 8 - seed: 42 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 - lr_scheduler_type: linear - lr_scheduler_warmup_ratio: 0.1 - num_epochs: 2 - mixed_precision_training: Native AMP ### Training results | Training Loss | Epoch | Step | Validation Loss | Precision | Recall | F1 | Accuracy | |:-------------:|:-----:|:----:|:---------------:|:---------:|:------:|:------:|:--------:| | 0.842 | 0.16 | 400 | 0.3530 | 0.5535 | 0.6872 | 0.6131 | 0.8927 | | 0.32 | 0.32 | 800 | 0.2800 | 0.6929 | 0.6749 | 0.6838 | 0.9190 | | 0.2928 | 0.48 | 1200 | 0.2438 | 0.7031 | 0.7661 | 0.7333 | 0.9301 | | 0.245 | 0.64 | 1600 | 0.2525 | 0.6959 | 0.7919 | 0.7408 | 0.9280 | | 0.2236 | 0.8 | 2000 | 0.2315 | 0.7441 | 0.7503 | 0.7472 | 0.9342 | | 0.2444 | 0.96 | 2400 | 0.2119 | 0.7719 | 0.7675 | 0.7697 | 0.9379 | | 0.1899 | 1.12 | 2800 | 0.2267 | 0.7531 | 0.8062 | 0.7788 | 0.9387 | | 0.1649 | 1.28 | 3200 | 0.2249 | 0.7519 | 0.8202 | 0.7846 | 0.9395 | | 0.1521 | 1.44 | 3600 | 0.2220 | 0.7778 | 0.8032 | 0.7903 | 0.9413 | | 0.1787 | 1.6 | 4000 | 0.2185 | 0.7879 | 0.7860 | 0.7869 | 0.9417 | | 0.146 | 1.76 | 4400 | 0.2134 | 0.7721 | 0.8128 | 0.7919 | 0.9416 | | 0.1557 | 1.92 | 4800 | 0.2111 | 0.7857 | 0.8101 | 0.7977 | 0.9429 | ### Framework versions - Transformers 4.27.4 - Pytorch 2.0.0+cu118 - Datasets 2.11.0 - Tokenizers 0.13.3
3,083
[ [ -0.0394287109375, -0.042938232421875, 0.0033721923828125, 0.00855255126953125, -0.0161590576171875, -0.025726318359375, -0.01068115234375, -0.0155181884765625, 0.0307769775390625, 0.0230865478515625, -0.0479736328125, -0.04931640625, -0.045989990234375, -0.0...
MohammedEltoum/dqn-SpaceInvadersNoFrameskip-v4
2023-04-12T03:25:35.000Z
[ "stable-baselines3", "SpaceInvadersNoFrameskip-v4", "deep-reinforcement-learning", "reinforcement-learning", "model-index", "region:us" ]
reinforcement-learning
MohammedEltoum
null
null
MohammedEltoum/dqn-SpaceInvadersNoFrameskip-v4
0
2
stable-baselines3
2023-04-12T03:24:49
--- library_name: stable-baselines3 tags: - SpaceInvadersNoFrameskip-v4 - deep-reinforcement-learning - reinforcement-learning - stable-baselines3 model-index: - name: DQN results: - task: type: reinforcement-learning name: reinforcement-learning dataset: name: SpaceInvadersNoFrameskip-v4 type: SpaceInvadersNoFrameskip-v4 metrics: - type: mean_reward value: 584.00 +/- 104.33 name: mean_reward verified: false --- # **DQN** Agent playing **SpaceInvadersNoFrameskip-v4** This is a trained model of a **DQN** agent playing **SpaceInvadersNoFrameskip-v4** using the [stable-baselines3 library](https://github.com/DLR-RM/stable-baselines3) and the [RL Zoo](https://github.com/DLR-RM/rl-baselines3-zoo). The RL Zoo is a training framework for Stable Baselines3 reinforcement learning agents, with hyperparameter optimization and pre-trained agents included. ## Usage (with SB3 RL Zoo) RL Zoo: https://github.com/DLR-RM/rl-baselines3-zoo<br/> SB3: https://github.com/DLR-RM/stable-baselines3<br/> SB3 Contrib: https://github.com/Stable-Baselines-Team/stable-baselines3-contrib Install the RL Zoo (with SB3 and SB3-Contrib): ```bash pip install rl_zoo3 ``` ``` # Download model and save it into the logs/ folder python -m rl_zoo3.load_from_hub --algo dqn --env SpaceInvadersNoFrameskip-v4 -orga MohammedEltoum -f logs/ python -m rl_zoo3.enjoy --algo dqn --env SpaceInvadersNoFrameskip-v4 -f logs/ ``` If you installed the RL Zoo3 via pip (`pip install rl_zoo3`), from anywhere you can do: ``` python -m rl_zoo3.load_from_hub --algo dqn --env SpaceInvadersNoFrameskip-v4 -orga MohammedEltoum -f logs/ python -m rl_zoo3.enjoy --algo dqn --env SpaceInvadersNoFrameskip-v4 -f logs/ ``` ## Training (with the RL Zoo) ``` python -m rl_zoo3.train --algo dqn --env SpaceInvadersNoFrameskip-v4 -f logs/ # Upload the model and generate video (when possible) python -m rl_zoo3.push_to_hub --algo dqn --env SpaceInvadersNoFrameskip-v4 -f logs/ -orga MohammedEltoum ``` ## Hyperparameters ```python OrderedDict([('batch_size', 32), ('buffer_size', 100000), ('env_wrapper', ['stable_baselines3.common.atari_wrappers.AtariWrapper']), ('exploration_final_eps', 0.01), ('exploration_fraction', 0.1), ('frame_stack', 4), ('gradient_steps', 1), ('learning_rate', 0.0001), ('learning_starts', 100000), ('n_timesteps', 1000000.0), ('optimize_memory_usage', False), ('policy', 'CnnPolicy'), ('target_update_interval', 1000), ('train_freq', 4), ('normalize', False)]) ```
2,709
[ [ -0.042083740234375, -0.036346435546875, 0.0210113525390625, 0.02447509765625, -0.01100921630859375, -0.0159912109375, 0.0131988525390625, -0.013397216796875, 0.0124053955078125, 0.0249481201171875, -0.07000732421875, -0.036529541015625, -0.028656005859375, -...
erickdp/fine-tuning-albert-tiny-041123
2023-04-12T05:54:04.000Z
[ "transformers", "pytorch", "albert", "text-classification", "generated_from_trainer", "endpoints_compatible", "region:us" ]
text-classification
erickdp
null
null
erickdp/fine-tuning-albert-tiny-041123
0
2
transformers
2023-04-12T03:31:04
--- tags: - generated_from_trainer metrics: - precision - f1 - recall - accuracy model-index: - name: fine-tuning-albert-tiny-041123 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. --> # fine-tuning-albert-tiny-041123 This model is a fine-tuned version of [dccuchile/albert-tiny-spanish](https://huggingface.co/dccuchile/albert-tiny-spanish) on the None dataset. It achieves the following results on the evaluation set: - Loss: 1.2027 - Precision: 0.1111 - F1: 0.1667 - Recall: 0.3333 - Accuracy: 0.3333 ## 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: 8 ### Training results | Training Loss | Epoch | Step | Validation Loss | Precision | F1 | Recall | Accuracy | |:-------------:|:-----:|:-----:|:---------------:|:---------:|:------:|:------:|:--------:| | 1.034 | 1.0 | 1304 | 1.2027 | 0.1111 | 0.1667 | 0.3333 | 0.3333 | | 1.0266 | 2.0 | 2608 | 1.1847 | 0.1111 | 0.1667 | 0.3333 | 0.3333 | | 1.0248 | 3.0 | 3912 | 1.1969 | 0.1111 | 0.1667 | 0.3333 | 0.3333 | | 1.0317 | 4.0 | 5216 | 1.2050 | 0.1111 | 0.1667 | 0.3333 | 0.3333 | | 1.0285 | 5.0 | 6520 | 1.1994 | 0.1111 | 0.1667 | 0.3333 | 0.3333 | | 1.0281 | 6.0 | 7824 | 1.1928 | 0.1111 | 0.1667 | 0.3333 | 0.3333 | | 1.0216 | 7.0 | 9128 | 1.2110 | 0.1111 | 0.1667 | 0.3333 | 0.3333 | | 1.0268 | 8.0 | 10432 | 1.2035 | 0.1111 | 0.1667 | 0.3333 | 0.3333 | ### Framework versions - Transformers 4.27.4 - Pytorch 1.13.1+cu116 - Datasets 2.11.0 - Tokenizers 0.13.3
2,177
[ [ -0.0469970703125, -0.037384033203125, 0.01526641845703125, 0.00885772705078125, -0.01343536376953125, -0.024566650390625, -0.01345062255859375, -0.01580810546875, 0.01557159423828125, 0.0166473388671875, -0.05206298828125, -0.04693603515625, -0.04345703125, ...
marice/distilbert-base-uncased-finetuned-clinc
2023-04-12T03:43:09.000Z
[ "transformers", "pytorch", "tensorboard", "distilbert", "text-classification", "generated_from_trainer", "dataset:clinc_oos", "license:apache-2.0", "model-index", "endpoints_compatible", "region:us" ]
text-classification
marice
null
null
marice/distilbert-base-uncased-finetuned-clinc
0
2
transformers
2023-04-12T03:35:29
--- license: apache-2.0 tags: - generated_from_trainer datasets: - clinc_oos metrics: - accuracy model-index: - name: distilbert-base-uncased-finetuned-clinc results: - task: name: Text Classification type: text-classification dataset: name: clinc_oos type: clinc_oos args: plus metrics: - name: Accuracy type: accuracy value: 0.9180645161290323 --- <!-- 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 the clinc_oos dataset. It achieves the following results on the evaluation set: - Loss: 0.7720 - Accuracy: 0.9181 ## 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: 5 ### Training results | Training Loss | Epoch | Step | Validation Loss | Accuracy | |:-------------:|:-----:|:----:|:---------------:|:--------:| | 4.2896 | 1.0 | 318 | 3.2887 | 0.7419 | | 2.6282 | 2.0 | 636 | 1.8753 | 0.8371 | | 1.548 | 3.0 | 954 | 1.1570 | 0.8961 | | 1.0148 | 4.0 | 1272 | 0.8573 | 0.9129 | | 0.7952 | 5.0 | 1590 | 0.7720 | 0.9181 | ### Framework versions - Transformers 4.11.3 - Pytorch 2.0.0+cu118 - Datasets 1.16.1 - Tokenizers 0.10.3
1,889
[ [ -0.034820556640625, -0.041046142578125, 0.012603759765625, 0.0071868896484375, -0.0266265869140625, -0.024688720703125, -0.01287841796875, -0.0084381103515625, 0.003124237060546875, 0.0218963623046875, -0.046478271484375, -0.048553466796875, -0.05841064453125, ...
rwang5688/distilgpt2-finetuned-wikitext2-pt
2023-10-13T21:37:20.000Z
[ "transformers", "pytorch", "tensorboard", "gpt2", "text-generation", "generated_from_trainer", "license:apache-2.0", "endpoints_compatible", "text-generation-inference", "region:us" ]
text-generation
rwang5688
null
null
rwang5688/distilgpt2-finetuned-wikitext2-pt
1
2
transformers
2023-04-12T04:23:56
--- license: apache-2.0 base_model: distilgpt2 tags: - generated_from_trainer model-index: - name: distilgpt2-finetuned-wikitext2-pt 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. --> # distilgpt2-finetuned-wikitext2-pt This model is a fine-tuned version of [distilgpt2](https://huggingface.co/distilgpt2) on the None dataset. It achieves the following results on the evaluation set: - Loss: 3.6429 ## 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: 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: 3.0 ### Training results | Training Loss | Epoch | Step | Validation Loss | |:-------------:|:-----:|:----:|:---------------:| | 3.7569 | 1.0 | 2334 | 3.6671 | | 3.6413 | 2.0 | 4668 | 3.6477 | | 3.596 | 3.0 | 7002 | 3.6429 | ### Framework versions - Transformers 4.34.0 - Pytorch 2.0.1+cu118 - Datasets 2.14.5 - Tokenizers 0.14.1
1,397
[ [ -0.0345458984375, -0.04241943359375, 0.013427734375, 0.0149993896484375, -0.0296478271484375, -0.031982421875, -0.00453948974609375, -0.0099029541015625, -0.00899505615234375, 0.01401519775390625, -0.0579833984375, -0.026458740234375, -0.05963134765625, -0.0...
Sigwang/distilbert-base-uncased-finetuned-emotion
2023-04-14T01:51:29.000Z
[ "transformers", "pytorch", "tensorboard", "distilbert", "text-classification", "generated_from_trainer", "dataset:emotion", "license:apache-2.0", "model-index", "endpoints_compatible", "region:us" ]
text-classification
Sigwang
null
null
Sigwang/distilbert-base-uncased-finetuned-emotion
0
2
transformers
2023-04-12T04:33:05
--- 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 config: split split: validation args: split metrics: - name: Accuracy type: accuracy value: 0.926 - name: F1 type: f1 value: 0.9261570669458271 --- <!-- 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: 0.2262 - Accuracy: 0.926 - F1: 0.9262 ## 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 | |:-------------:|:-----:|:----:|:---------------:|:--------:|:------:| | 0.837 | 1.0 | 250 | 0.3302 | 0.9015 | 0.8980 | | 0.2559 | 2.0 | 500 | 0.2262 | 0.926 | 0.9262 | ### Framework versions - Transformers 4.28.0 - Pytorch 2.0.0+cu118 - Datasets 2.11.0 - Tokenizers 0.13.3
1,846
[ [ -0.037384033203125, -0.041473388671875, 0.0146331787109375, 0.0218353271484375, -0.026458740234375, -0.0192413330078125, -0.0135040283203125, -0.0085601806640625, 0.01023101806640625, 0.00748443603515625, -0.056365966796875, -0.052093505859375, -0.06015014648437...
nes74/distilbert-base-uncased-finetuned-emotion
2023-06-03T02:56:01.000Z
[ "transformers", "pytorch", "tensorboard", "distilbert", "text-classification", "generated_from_trainer", "dataset:emotion", "license:apache-2.0", "model-index", "endpoints_compatible", "region:us" ]
text-classification
nes74
null
null
nes74/distilbert-base-uncased-finetuned-emotion
0
2
transformers
2023-04-12T05:43:37
--- 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 config: split split: validation args: split metrics: - name: Accuracy type: accuracy value: 0.926 - name: F1 type: f1 value: 0.9260997886540973 --- <!-- 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: 0.2210 - Accuracy: 0.926 - F1: 0.9261 ## 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 | |:-------------:|:-----:|:----:|:---------------:|:--------:|:------:| | 0.8246 | 1.0 | 250 | 0.3126 | 0.909 | 0.9075 | | 0.2525 | 2.0 | 500 | 0.2210 | 0.926 | 0.9261 | ### Framework versions - Transformers 4.29.2 - Pytorch 2.0.1 - Datasets 2.12.0 - Tokenizers 0.13.2
1,840
[ [ -0.03765869140625, -0.040313720703125, 0.0145721435546875, 0.02215576171875, -0.0264129638671875, -0.01995849609375, -0.012786865234375, -0.008636474609375, 0.0099334716796875, 0.00823974609375, -0.055938720703125, -0.05206298828125, -0.059600830078125, -0.0...
marice/distilbert-base-uncased-distilled-clinc
2023-04-12T06:09:45.000Z
[ "transformers", "pytorch", "distilbert", "text-classification", "generated_from_trainer", "dataset:clinc_oos", "license:apache-2.0", "model-index", "endpoints_compatible", "region:us" ]
text-classification
marice
null
null
marice/distilbert-base-uncased-distilled-clinc
0
2
transformers
2023-04-12T05:57:17
--- license: apache-2.0 tags: - generated_from_trainer datasets: - clinc_oos metrics: - accuracy model-index: - name: distilbert-base-uncased-distilled-clinc results: - task: name: Text Classification type: text-classification dataset: name: clinc_oos type: clinc_oos args: plus metrics: - name: Accuracy type: accuracy value: 0.9387096774193548 --- <!-- 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-distilled-clinc This model is a fine-tuned version of [distilbert-base-uncased](https://huggingface.co/distilbert-base-uncased) on the clinc_oos dataset. It achieves the following results on the evaluation set: - Loss: 0.0878 - Accuracy: 0.9387 ## 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: 10 ### Training results | Training Loss | Epoch | Step | Validation Loss | Accuracy | |:-------------:|:-----:|:----:|:---------------:|:--------:| | 1.0369 | 1.0 | 318 | 0.5902 | 0.6987 | | 0.4468 | 2.0 | 636 | 0.2434 | 0.8606 | | 0.2204 | 3.0 | 954 | 0.1412 | 0.9113 | | 0.1478 | 4.0 | 1272 | 0.1121 | 0.9252 | | 0.1206 | 5.0 | 1590 | 0.1010 | 0.93 | | 0.1086 | 6.0 | 1908 | 0.0947 | 0.9345 | | 0.1009 | 7.0 | 2226 | 0.0916 | 0.9368 | | 0.0966 | 8.0 | 2544 | 0.0896 | 0.9381 | | 0.0939 | 9.0 | 2862 | 0.0881 | 0.9390 | | 0.0928 | 10.0 | 3180 | 0.0878 | 0.9387 | ### Framework versions - Transformers 4.11.3 - Pytorch 2.0.0+cu118 - Datasets 1.16.1 - Tokenizers 0.10.3
2,200
[ [ -0.034637451171875, -0.038238525390625, 0.0160980224609375, 0.00658416748046875, -0.0223236083984375, -0.01678466796875, -0.0078125, -0.0041656494140625, 0.01129150390625, 0.022064208984375, -0.04412841796875, -0.049652099609375, -0.0601806640625, -0.0095748...
Jcfranco/distilbert-base-uncased-finetuned-sst2
2023-04-12T11:08:31.000Z
[ "transformers", "pytorch", "tensorboard", "distilbert", "text-classification", "generated_from_trainer", "dataset:glue", "license:apache-2.0", "model-index", "endpoints_compatible", "region:us" ]
text-classification
Jcfranco
null
null
Jcfranco/distilbert-base-uncased-finetuned-sst2
0
2
transformers
2023-04-12T06:25:59
--- license: apache-2.0 tags: - generated_from_trainer datasets: - glue metrics: - accuracy model-index: - name: distilbert-base-uncased-finetuned-sst2 results: - task: name: Text Classification type: text-classification dataset: name: glue type: glue config: sst2 split: validation args: sst2 metrics: - name: Accuracy type: accuracy value: 0.908256880733945 --- <!-- 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-sst2 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: 0.3078 - Accuracy: 0.9083 ## 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: 5 ### Training results | Training Loss | Epoch | Step | Validation Loss | Accuracy | |:-------------:|:-----:|:----:|:---------------:|:--------:| | No log | 1.0 | 211 | 0.3078 | 0.9083 | | No log | 2.0 | 422 | 0.4370 | 0.8968 | | 0.0968 | 3.0 | 633 | 0.4457 | 0.9002 | | 0.0968 | 4.0 | 844 | 0.4723 | 0.9048 | | 0.0259 | 5.0 | 1055 | 0.4991 | 0.9014 | ### Framework versions - Transformers 4.27.4 - Pytorch 2.0.0+cu118 - Datasets 2.11.0 - Tokenizers 0.13.3
1,909
[ [ -0.0188751220703125, -0.04852294921875, 0.01419830322265625, 0.01242828369140625, -0.0307769775390625, -0.0157623291015625, -0.0093536376953125, -0.0040283203125, 0.005950927734375, 0.01432037353515625, -0.047210693359375, -0.038055419921875, -0.06085205078125, ...
fathyshalab/massive-ar-SA
2023-04-12T11:08:05.000Z
[ "sentence-transformers", "pytorch", "bert", "setfit", "text-classification", "arxiv:2209.11055", "license:apache-2.0", "region:us" ]
text-classification
fathyshalab
null
null
fathyshalab/massive-ar-SA
0
2
sentence-transformers
2023-04-12T11:07:38
--- license: apache-2.0 tags: - setfit - sentence-transformers - text-classification pipeline_tag: text-classification --- # fathyshalab/massive-ar-SA This is a [SetFit model](https://github.com/huggingface/setfit) that can be used for text classification. The model has been trained using an efficient few-shot learning technique that involves: 1. Fine-tuning a [Sentence Transformer](https://www.sbert.net) with contrastive learning. 2. Training a classification head with features from the fine-tuned Sentence Transformer. ## Usage To use this model for inference, first install the SetFit library: ```bash python -m pip install setfit ``` You can then run inference as follows: ```python from setfit import SetFitModel # Download from Hub and run inference model = SetFitModel.from_pretrained("fathyshalab/massive-ar-SA") # Run inference preds = model(["i loved the spiderman movie!", "pineapple on pizza is the worst 🤮"]) ``` ## BibTeX entry and citation info ```bibtex @article{https://doi.org/10.48550/arxiv.2209.11055, doi = {10.48550/ARXIV.2209.11055}, url = {https://arxiv.org/abs/2209.11055}, author = {Tunstall, Lewis and Reimers, Nils and Jo, Unso Eun Seo and Bates, Luke and Korat, Daniel and Wasserblat, Moshe and Pereg, Oren}, keywords = {Computation and Language (cs.CL), FOS: Computer and information sciences, FOS: Computer and information sciences}, title = {Efficient Few-Shot Learning Without Prompts}, publisher = {arXiv}, year = {2022}, copyright = {Creative Commons Attribution 4.0 International} } ```
1,539
[ [ -0.0061187744140625, -0.056060791015625, 0.0242919921875, -0.016693115234375, -0.01235198974609375, -0.01354217529296875, -0.0154266357421875, -0.01202392578125, 0.005817413330078125, 0.03143310546875, -0.0302734375, -0.0213775634765625, -0.046295166015625, ...
SakuraKnight/Yelp-Rating-Prediction
2023-04-12T18:23:48.000Z
[ "transformers", "pytorch", "bert", "text-classification", "license:mit", "endpoints_compatible", "region:us" ]
text-classification
SakuraKnight
null
null
SakuraKnight/Yelp-Rating-Prediction
0
2
transformers
2023-04-12T12:36:28
--- license: mit --- A Demo BERT classification model Trained on (Part of) Yelp Dataset Photo2Text model: ydshieh/vit-gpt2-coco-en Expected / Standard Input: ``` [CLS] Business Name [SEP] Address [SEP] City [SEP] Photo2Text Outputs ... ``` Example: ``` [CLS] Paws The Cat Cafe [SEP] 10588 109 Street [SEP] Edmonton [SEP] A cup of coffee ``` Expected Output: 5
366
[ [ -0.0004227161407470703, -0.04486083984375, 0.043243408203125, 0.0093536376953125, -0.0260162353515625, -0.01482391357421875, 0.00907135009765625, -0.04180908203125, 0.024658203125, 0.04193115234375, -0.052490234375, -0.04632568359375, -0.033966064453125, 0.0...
ArunaSaraswathy/pii_new_model
2023-04-12T14:18:32.000Z
[ "transformers", "pytorch", "distilbert", "token-classification", "generated_from_trainer", "license:apache-2.0", "autotrain_compatible", "endpoints_compatible", "region:us" ]
token-classification
ArunaSaraswathy
null
null
ArunaSaraswathy/pii_new_model
0
2
transformers
2023-04-12T14:08:40
--- license: apache-2.0 tags: - generated_from_trainer model-index: - name: results 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. --> # results This model is a fine-tuned version of [distilbert-base-cased](https://huggingface.co/distilbert-base-cased) on an unknown 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: 5e-05 - train_batch_size: 16 - eval_batch_size: 64 - seed: 42 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 - lr_scheduler_type: linear - lr_scheduler_warmup_steps: 500 - num_epochs: 3 ### Training results ### Framework versions - Transformers 4.24.0 - Pytorch 2.0.0 - Datasets 2.9.0 - Tokenizers 0.13.2
1,045
[ [ -0.033050537109375, -0.056610107421875, 0.012420654296875, 0.0171051025390625, -0.032684326171875, -0.01947021484375, -0.007049560546875, -0.00699615478515625, 0.0074310302734375, 0.021728515625, -0.0545654296875, -0.045196533203125, -0.05633544921875, -0.00...
bblackwell/distilbert-base-uncased-finetuned-cola
2023-04-25T14:26:10.000Z
[ "transformers", "pytorch", "tensorboard", "distilbert", "text-classification", "generated_from_trainer", "license:apache-2.0", "endpoints_compatible", "region:us" ]
text-classification
bblackwell
null
null
bblackwell/distilbert-base-uncased-finetuned-cola
0
2
transformers
2023-04-12T14:33:52
--- license: apache-2.0 tags: - generated_from_trainer metrics: - matthews_correlation model-index: - name: distilbert-base-uncased-finetuned-cola 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-cola 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: 0.1581 - Matthews Correlation: 0.8855 ## 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: 8.908469178483356e-05 - train_batch_size: 64 - eval_batch_size: 16 - seed: 31 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 - lr_scheduler_type: linear - num_epochs: 3 ### Training results | Training Loss | Epoch | Step | Validation Loss | Matthews Correlation | |:-------------:|:-----:|:----:|:---------------:|:--------------------:| | No log | 1.0 | 146 | 0.1384 | 0.8905 | | No log | 2.0 | 292 | 0.1361 | 0.8738 | | No log | 3.0 | 438 | 0.1581 | 0.8855 | ### Framework versions - Transformers 4.28.1 - Pytorch 2.0.0+cu118 - Datasets 2.11.0 - Tokenizers 0.13.3
1,606
[ [ -0.023101806640625, -0.051788330078125, 0.0158233642578125, 0.021087646484375, -0.0259246826171875, -0.0079803466796875, -0.007232666015625, -0.0084228515625, 0.0189208984375, 0.0166015625, -0.045440673828125, -0.0352783203125, -0.062164306640625, -0.0046882...
jorgeortizfuentes/spanish-spellchecker-flan-t5-base-wiki200000
2023-04-13T07:55:20.000Z
[ "transformers", "pytorch", "tensorboard", "t5", "text2text-generation", "generated_from_trainer", "license:apache-2.0", "autotrain_compatible", "endpoints_compatible", "text-generation-inference", "region:us" ]
text2text-generation
jorgeortizfuentes
null
null
jorgeortizfuentes/spanish-spellchecker-flan-t5-base-wiki200000
0
2
transformers
2023-04-12T15:03:57
--- license: apache-2.0 tags: - generated_from_trainer metrics: - bleu model-index: - name: spanish-spellchecker-flan-t5-base-wiki200000 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. --> # spanish-spellchecker-flan-t5-base-wiki200000 This model is a fine-tuned version of [google/flan-t5-base](https://huggingface.co/google/flan-t5-base) on the None dataset. It achieves the following results on the evaluation set: - Loss: 0.1471 - Bleu: 0.0 - Gen Len: 19.0 ## 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 - distributed_type: multi-GPU - num_devices: 2 - total_train_batch_size: 64 - total_eval_batch_size: 64 - 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 | Bleu | Gen Len | |:-------------:|:-----:|:-----:|:---------------:|:----:|:-------:| | 0.1876 | 1.0 | 9755 | 0.1550 | 0.0 | 19.0 | | 0.1768 | 2.0 | 19510 | 0.1471 | 0.0 | 19.0 | ### Framework versions - Transformers 4.26.1 - Pytorch 1.13.1+cu117 - Datasets 2.10.1 - Tokenizers 0.13.2
1,584
[ [ -0.02783203125, -0.036468505859375, 0.00994110107421875, 0.01425933837890625, -0.01134490966796875, -0.027679443359375, -0.0186767578125, -0.027069091796875, 0.01436614990234375, 0.0201416015625, -0.049102783203125, -0.0526123046875, -0.0546875, 0.0068969726...
GhifSmile/distilbert-base-uncased-PINA-dfnew
2023-04-12T18:42:43.000Z
[ "transformers", "pytorch", "tensorboard", "distilbert", "text-classification", "generated_from_trainer", "license:apache-2.0", "endpoints_compatible", "region:us" ]
text-classification
GhifSmile
null
null
GhifSmile/distilbert-base-uncased-PINA-dfnew
0
2
transformers
2023-04-12T15:33:09
--- license: apache-2.0 tags: - generated_from_trainer metrics: - accuracy - precision - recall model-index: - name: distilbert-base-uncased-PINA-dfnew 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-PINA-dfnew 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: 0.2599 - Accuracy: 0.9510 - Precision: 0.8737 - Recall: 0.8532 ## 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: 10 ### Training results | Training Loss | Epoch | Step | Validation Loss | Accuracy | Precision | Recall | |:-------------:|:-----:|:-----:|:---------------:|:--------:|:---------:|:------:| | 1.1798 | 1.0 | 1438 | 0.4320 | 0.9016 | 0.7777 | 0.7182 | | 0.2987 | 2.0 | 2876 | 0.2779 | 0.9369 | 0.8340 | 0.8270 | | 0.1579 | 3.0 | 4314 | 0.2608 | 0.9445 | 0.8374 | 0.8378 | | 0.0913 | 4.0 | 5752 | 0.2599 | 0.9510 | 0.8737 | 0.8532 | | 0.0547 | 5.0 | 7190 | 0.2716 | 0.9531 | 0.8893 | 0.8682 | | 0.0309 | 6.0 | 8628 | 0.2748 | 0.9531 | 0.8921 | 0.8750 | | 0.0174 | 7.0 | 10066 | 0.2860 | 0.9545 | 0.8966 | 0.8710 | | 0.01 | 8.0 | 11504 | 0.2972 | 0.9543 | 0.9087 | 0.8989 | | 0.0063 | 9.0 | 12942 | 0.3012 | 0.9536 | 0.9066 | 0.8967 | | 0.0044 | 10.0 | 14380 | 0.2978 | 0.9551 | 0.9108 | 0.8997 | ### Framework versions - Transformers 4.27.4 - Pytorch 2.0.0+cu118 - Datasets 2.11.0 - Tokenizers 0.13.3
2,253
[ [ -0.036468505859375, -0.041046142578125, 0.018096923828125, 0.011322021484375, -0.017669677734375, -0.01181793212890625, 0.00007522106170654297, -0.001987457275390625, 0.0262908935546875, 0.01904296875, -0.046600341796875, -0.050933837890625, -0.0550537109375, ...
AyoubChLin/XLMRoberta-large-bbc_news
2023-04-12T18:13:52.000Z
[ "transformers", "pytorch", "xlm-roberta", "text-classification", "autotrain", "en", "dataset:AyoubChLin/autotrain-data-anymodel_bbc", "dataset:SetFit/bbc-news", "license:apache-2.0", "co2_eq_emissions", "endpoints_compatible", "region:us" ]
text-classification
AyoubChLin
null
null
AyoubChLin/XLMRoberta-large-bbc_news
0
2
transformers
2023-04-12T16:32:48
--- tags: - autotrain - text-classification language: - en widget: - text: A new model offers an explanation for how the Galilean satellites formed around the solar system’s largest world. Konstantin Batygin did not set out to solve one of the solar system’s most puzzling mysteries when he went for a run up a hill in Nice, France. Dr. Batygin, a Caltech researcher datasets: - AyoubChLin/autotrain-data-anymodel_bbc - SetFit/bbc-news co2_eq_emissions: emissions: 2.359134715120443 license: apache-2.0 metrics: - accuracy pipeline_tag: text-classification --- # Model Trained Using AutoTrain - Problem type: Multi-class Classification - Model ID: 48900118383 - CO2 Emissions (in grams): 2.3591 ## Validation Metrics - Loss: 0.116 - Accuracy: 0.978 - Macro F1: 0.978 - Micro F1: 0.978 - Weighted F1: 0.978 - Macro Precision: 0.978 - Micro Precision: 0.978 - Weighted Precision: 0.978 - Macro Recall: 0.978 - Micro Recall: 0.978 - Weighted Recall: 0.978 ## 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/AyoubChLin/autotrain-anymodel_bbc-48900118383 ``` Or Python API: ``` from transformers import AutoModelForSequenceClassification, AutoTokenizer model = AutoModelForSequenceClassification.from_pretrained("AyoubChLin/autotrain-anymodel_bbc-48900118383", use_auth_token=True) tokenizer = AutoTokenizer.from_pretrained("AyoubChLin/autotrain-anymodel_bbc-48900118383", use_auth_token=True) inputs = tokenizer("I love AutoTrain", return_tensors="pt") outputs = model(**inputs) ```
1,666
[ [ -0.033538818359375, -0.0256195068359375, 0.00873565673828125, 0.01190948486328125, -0.0028438568115234375, 0.0009083747863769531, 0.0021877288818359375, -0.013275146484375, -0.00107574462890625, 0.01071929931640625, -0.05224609375, -0.031097412109375, -0.0574645...
Telstema/distilbert-base-uncased-finetuned-cola
2023-04-13T01:16:25.000Z
[ "transformers", "pytorch", "tensorboard", "distilbert", "text-classification", "generated_from_trainer", "dataset:glue", "license:apache-2.0", "model-index", "endpoints_compatible", "region:us" ]
text-classification
Telstema
null
null
Telstema/distilbert-base-uncased-finetuned-cola
0
2
transformers
2023-04-12T16:36:59
--- 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 config: cola split: validation args: cola metrics: - name: Matthews Correlation type: matthews_correlation value: 0.56217893832047 --- <!-- 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: 0.7649 - Matthews Correlation: 0.5622 ## 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 | |:-------------:|:-----:|:----:|:---------------:|:--------------------:| | 0.5218 | 1.0 | 535 | 0.5275 | 0.4033 | | 0.3492 | 2.0 | 1070 | 0.5052 | 0.4987 | | 0.2362 | 3.0 | 1605 | 0.5527 | 0.5382 | | 0.1763 | 4.0 | 2140 | 0.7443 | 0.5378 | | 0.1212 | 5.0 | 2675 | 0.7649 | 0.5622 | ### Framework versions - Transformers 4.27.4 - Pytorch 2.0.0+cu118 - Datasets 2.11.0 - Tokenizers 0.13.3
2,040
[ [ -0.02227783203125, -0.04998779296875, 0.01131439208984375, 0.0186767578125, -0.0230255126953125, -0.007778167724609375, -0.00502777099609375, -0.0036029815673828125, 0.0235137939453125, 0.0097808837890625, -0.0447998046875, -0.03509521484375, -0.06268310546875, ...
OMARS200/SpaceInvadersNoFrameskip-v4
2023-04-12T16:53:03.000Z
[ "stable-baselines3", "SpaceInvadersNoFrameskip-v4", "deep-reinforcement-learning", "reinforcement-learning", "model-index", "region:us" ]
reinforcement-learning
OMARS200
null
null
OMARS200/SpaceInvadersNoFrameskip-v4
0
2
stable-baselines3
2023-04-12T16:52:27
--- library_name: stable-baselines3 tags: - SpaceInvadersNoFrameskip-v4 - deep-reinforcement-learning - reinforcement-learning - stable-baselines3 model-index: - name: DQN results: - task: type: reinforcement-learning name: reinforcement-learning dataset: name: SpaceInvadersNoFrameskip-v4 type: SpaceInvadersNoFrameskip-v4 metrics: - type: mean_reward value: 587.50 +/- 217.47 name: mean_reward verified: false --- # **DQN** Agent playing **SpaceInvadersNoFrameskip-v4** This is a trained model of a **DQN** agent playing **SpaceInvadersNoFrameskip-v4** using the [stable-baselines3 library](https://github.com/DLR-RM/stable-baselines3) and the [RL Zoo](https://github.com/DLR-RM/rl-baselines3-zoo). The RL Zoo is a training framework for Stable Baselines3 reinforcement learning agents, with hyperparameter optimization and pre-trained agents included. ## Usage (with SB3 RL Zoo) RL Zoo: https://github.com/DLR-RM/rl-baselines3-zoo<br/> SB3: https://github.com/DLR-RM/stable-baselines3<br/> SB3 Contrib: https://github.com/Stable-Baselines-Team/stable-baselines3-contrib Install the RL Zoo (with SB3 and SB3-Contrib): ```bash pip install rl_zoo3 ``` ``` # Download model and save it into the logs/ folder python -m rl_zoo3.load_from_hub --algo dqn --env SpaceInvadersNoFrameskip-v4 -orga OMARS200 -f logs/ python -m rl_zoo3.enjoy --algo dqn --env SpaceInvadersNoFrameskip-v4 -f logs/ ``` If you installed the RL Zoo3 via pip (`pip install rl_zoo3`), from anywhere you can do: ``` python -m rl_zoo3.load_from_hub --algo dqn --env SpaceInvadersNoFrameskip-v4 -orga OMARS200 -f logs/ python -m rl_zoo3.enjoy --algo dqn --env SpaceInvadersNoFrameskip-v4 -f logs/ ``` ## Training (with the RL Zoo) ``` python -m rl_zoo3.train --algo dqn --env SpaceInvadersNoFrameskip-v4 -f logs/ # Upload the model and generate video (when possible) python -m rl_zoo3.push_to_hub --algo dqn --env SpaceInvadersNoFrameskip-v4 -f logs/ -orga OMARS200 ``` ## Hyperparameters ```python OrderedDict([('batch_size', 32), ('buffer_size', 100000), ('env_wrapper', ['stable_baselines3.common.atari_wrappers.AtariWrapper']), ('exploration_final_eps', 0.01), ('exploration_fraction', 0.1), ('frame_stack', 4), ('gradient_steps', 1), ('learning_rate', 0.0001), ('learning_starts', 100000), ('n_timesteps', 1000000.0), ('optimize_memory_usage', False), ('policy', 'CnnPolicy'), ('target_update_interval', 1000), ('train_freq', 4), ('normalize', False)]) ```
2,691
[ [ -0.041412353515625, -0.0361328125, 0.0218963623046875, 0.0251617431640625, -0.009979248046875, -0.0164794921875, 0.01264190673828125, -0.01331329345703125, 0.0131988525390625, 0.02447509765625, -0.07049560546875, -0.035888671875, -0.027557373046875, -0.00504...
asubiabre/ppo-PyramidsTraining
2023-04-12T17:32:41.000Z
[ "ml-agents", "tensorboard", "onnx", "Pyramids", "deep-reinforcement-learning", "reinforcement-learning", "ML-Agents-Pyramids", "region:us" ]
reinforcement-learning
asubiabre
null
null
asubiabre/ppo-PyramidsTraining
0
2
ml-agents
2023-04-12T17:32:35
--- library_name: ml-agents tags: - Pyramids - deep-reinforcement-learning - reinforcement-learning - ML-Agents-Pyramids --- # **ppo** Agent playing **Pyramids** This is a trained model of a **ppo** agent playing **Pyramids** using the [Unity ML-Agents Library](https://github.com/Unity-Technologies/ml-agents). ## Usage (with ML-Agents) The Documentation: https://github.com/huggingface/ml-agents#get-started We wrote a complete tutorial to learn to train your first agent using ML-Agents and publish it to the Hub: ### Resume the training ``` mlagents-learn <your_configuration_file_path.yaml> --run-id=<run_id> --resume ``` ### Watch your Agent play You can watch your agent **playing directly in your browser:**. 1. Go to https://huggingface.co/spaces/unity/ML-Agents-Pyramids 2. Step 1: Find your model_id: asubiabre/ppo-PyramidsTraining 3. Step 2: Select your *.nn /*.onnx file 4. Click on Watch the agent play 👀
960
[ [ -0.026397705078125, -0.018798828125, -0.0010633468627929688, 0.0274200439453125, -0.00982666015625, 0.0062103271484375, 0.0273895263671875, -0.002490997314453125, 0.034820556640625, 0.0357666015625, -0.036285400390625, -0.052215576171875, -0.035400390625, -0...
AyoubChLin/DistilRoberta-bbc_news
2023-04-12T21:49:32.000Z
[ "transformers", "pytorch", "roberta", "text-classification", "autotrain", "unk", "dataset:AyoubChLin/autotrain-data-distilroberta-bbc_news", "dataset:SetFit/bbc-news", "license:apache-2.0", "endpoints_compatible", "region:us" ]
text-classification
AyoubChLin
null
null
AyoubChLin/DistilRoberta-bbc_news
0
2
transformers
2023-04-12T18:54:42
--- tags: - autotrain - text-classification language: - unk widget: - text: I love AutoTrain 🤗 datasets: - AyoubChLin/autotrain-data-distilroberta-bbc_news - SetFit/bbc-news license: apache-2.0 metrics: - accuracy pipeline_tag: text-classification --- # Model Trained Using AutoTrain - Problem type: Multi-class Classification - Model ID: 48937118428 - CO2 Emissions (in grams): 0.6873 ## Validation Metrics - Loss: 0.063 - Accuracy: 0.985 - Macro F1: 0.984 - Micro F1: 0.985 - Weighted F1: 0.985 - Macro Precision: 0.984 - Micro Precision: 0.985 - Weighted Precision: 0.985 - Macro Recall: 0.985 - Micro Recall: 0.985 - Weighted Recall: 0.985 ## 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/AyoubChLin/autotrain-distilroberta-bbc_news-48937118428 ``` Or Python API: ``` from transformers import AutoModelForSequenceClassification, AutoTokenizer model = AutoModelForSequenceClassification.from_pretrained("AyoubChLin/autotrain-distilroberta-bbc_news-48937118428", use_auth_token=True) tokenizer = AutoTokenizer.from_pretrained("AyoubChLin/autotrain-distilroberta-bbc_news-48937118428", use_auth_token=True) inputs = tokenizer("I love AutoTrain", return_tensors="pt") outputs = model(**inputs) ```
1,385
[ [ -0.0310211181640625, -0.027984619140625, 0.007587432861328125, 0.01053619384765625, -0.0030307769775390625, 0.006420135498046875, -0.0017786026000976562, -0.0133819580078125, -0.00475311279296875, 0.004486083984375, -0.0458984375, -0.03326416015625, -0.058197021...
AyoubChLin/delberta_large_bbc_news
2023-04-12T19:25:14.000Z
[ "transformers", "pytorch", "deberta-v2", "text-classification", "autotrain", "en", "dataset:AyoubChLin/autotrain-data-delberta-large", "dataset:SetFit/bbc-news", "license:apache-2.0", "co2_eq_emissions", "endpoints_compatible", "region:us" ]
text-classification
AyoubChLin
null
null
AyoubChLin/delberta_large_bbc_news
0
2
transformers
2023-04-12T18:59:38
--- tags: - autotrain - text-classification language: - en widget: - text: I love AutoTrain 🤗 datasets: - AyoubChLin/autotrain-data-delberta-large - SetFit/bbc-news co2_eq_emissions: emissions: 4.083685268664441 license: apache-2.0 metrics: - accuracy pipeline_tag: text-classification --- # Model Trained Using AutoTrain - Problem type: Multi-class Classification - Model ID: 48938118433 - CO2 Emissions (in grams): 4.0837 ## Validation Metrics - Loss: 0.130 - Accuracy: 0.980 - Macro F1: 0.980 - Micro F1: 0.980 - Weighted F1: 0.980 - Macro Precision: 0.980 - Micro Precision: 0.980 - Weighted Precision: 0.980 - Macro Recall: 0.980 - Micro Recall: 0.980 - Weighted Recall: 0.980 ## 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/AyoubChLin/autotrain-delberta-large-48938118433 ``` Or Python API: ``` from transformers import AutoModelForSequenceClassification, AutoTokenizer model = AutoModelForSequenceClassification.from_pretrained("AyoubChLin/autotrain-delberta-large-48938118433", use_auth_token=True) tokenizer = AutoTokenizer.from_pretrained("AyoubChLin/autotrain-delberta-large-48938118433", use_auth_token=True) inputs = tokenizer("I love AutoTrain", return_tensors="pt") outputs = model(**inputs) ```
1,402
[ [ -0.034515380859375, -0.0261993408203125, 0.0123291015625, 0.01351165771484375, -0.00524139404296875, 0.0036334991455078125, 0.00125885009765625, -0.0153350830078125, 0.004543304443359375, 0.00603485107421875, -0.050140380859375, -0.03363037109375, -0.05859375, ...
AyoubChLin/Albert-bbc-news
2023-04-12T21:39:16.000Z
[ "transformers", "pytorch", "albert", "text-classification", "autotrain", "en", "dataset:AyoubChLin/autotrain-data-albert-bbc-news", "dataset:SetFit/bbc-news", "license:apache-2.0", "co2_eq_emissions", "endpoints_compatible", "region:us" ]
text-classification
AyoubChLin
null
null
AyoubChLin/Albert-bbc-news
0
2
transformers
2023-04-12T19:00:44
--- tags: - autotrain - text-classification language: - en widget: - text: I love AutoTrain 🤗 datasets: - AyoubChLin/autotrain-data-albert-bbc-news - SetFit/bbc-news co2_eq_emissions: emissions: 13.344689233410659 license: apache-2.0 metrics: - accuracy pipeline_tag: text-classification --- # Model Trained Using AutoTrain - Problem type: Multi-class Classification - Model ID: 48939118438 - CO2 Emissions (in grams): 13.3447 ## Validation Metrics - Loss: 0.103 - Accuracy: 0.978 - Macro F1: 0.978 - Micro F1: 0.978 - Weighted F1: 0.978 - Macro Precision: 0.977 - Micro Precision: 0.978 - Weighted Precision: 0.978 - Macro Recall: 0.978 - Micro Recall: 0.978 - Weighted Recall: 0.978 ## 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/AyoubChLin/autotrain-albert-bbc-news-48939118438 ``` Or Python API: ``` from transformers import AutoModelForSequenceClassification, AutoTokenizer model = AutoModelForSequenceClassification.from_pretrained("AyoubChLin/autotrain-albert-bbc-news-48939118438", use_auth_token=True) tokenizer = AutoTokenizer.from_pretrained("AyoubChLin/autotrain-albert-bbc-news-48939118438", use_auth_token=True) inputs = tokenizer("I love AutoTrain", return_tensors="pt") outputs = model(**inputs) ```
1,407
[ [ -0.0343017578125, -0.0265045166015625, 0.0084228515625, 0.01398468017578125, -0.0028095245361328125, 0.004566192626953125, 0.000033855438232421875, -0.01181793212890625, -0.0031337738037109375, 0.00844573974609375, -0.049407958984375, -0.0313720703125, -0.057678...
AyoubChLin/roberta-large-bbc_news
2023-08-08T20:29:24.000Z
[ "transformers", "pytorch", "safetensors", "roberta", "text-classification", "autotrain", "unk", "dataset:AyoubChLin/autotrain-data-roberta-large-bbc_news", "co2_eq_emissions", "endpoints_compatible", "region:us" ]
text-classification
AyoubChLin
null
null
AyoubChLin/roberta-large-bbc_news
0
2
transformers
2023-04-12T19:09:36
--- tags: - autotrain - text-classification language: - unk widget: - text: "I love AutoTrain 🤗" datasets: - AyoubChLin/autotrain-data-roberta-large-bbc_news co2_eq_emissions: emissions: 1.9843929651071104 --- # Model Trained Using AutoTrain - Problem type: Multi-class Classification - Model ID: 48943118458 - CO2 Emissions (in grams): 1.9844 ## Validation Metrics - Loss: 0.062 - Accuracy: 0.991 - Macro F1: 0.991 - Micro F1: 0.991 - Weighted F1: 0.991 - Macro Precision: 0.991 - Micro Precision: 0.991 - Weighted Precision: 0.991 - Macro Recall: 0.992 - Micro Recall: 0.991 - Weighted Recall: 0.991 ## 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/AyoubChLin/autotrain-roberta-large-bbc_news-48943118458 ``` Or Python API: ``` from transformers import AutoModelForSequenceClassification, AutoTokenizer model = AutoModelForSequenceClassification.from_pretrained("AyoubChLin/autotrain-roberta-large-bbc_news-48943118458", use_auth_token=True) tokenizer = AutoTokenizer.from_pretrained("AyoubChLin/autotrain-roberta-large-bbc_news-48943118458", use_auth_token=True) inputs = tokenizer("I love AutoTrain", return_tensors="pt") outputs = model(**inputs) ```
1,345
[ [ -0.031005859375, -0.0267181396484375, 0.00868988037109375, 0.0110931396484375, -0.0017299652099609375, 0.004180908203125, -0.0026988983154296875, -0.0135498046875, -0.002231597900390625, 0.00589752197265625, -0.04718017578125, -0.033935546875, -0.057281494140625...
bblackwell/distilbert-base-uncased-finetuned-cola-Christianity
2023-04-25T19:04:25.000Z
[ "transformers", "pytorch", "tensorboard", "distilbert", "text-classification", "generated_from_trainer", "license:apache-2.0", "endpoints_compatible", "region:us" ]
text-classification
bblackwell
null
null
bblackwell/distilbert-base-uncased-finetuned-cola-Christianity
0
2
transformers
2023-04-12T20:00:52
--- license: apache-2.0 tags: - generated_from_trainer metrics: - matthews_correlation model-index: - name: distilbert-base-uncased-finetuned-cola-Christianity 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-cola-Christianity 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: 0.2016 - Matthews Correlation: 0.8950 ## 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 | |:-------------:|:-----:|:----:|:---------------:|:--------------------:| | 0.2197 | 1.0 | 615 | 0.1455 | 0.8815 | | 0.13 | 2.0 | 1230 | 0.1453 | 0.8853 | | 0.0859 | 3.0 | 1845 | 0.1854 | 0.8879 | | 0.0511 | 4.0 | 2460 | 0.2016 | 0.8950 | | 0.0183 | 5.0 | 3075 | 0.2132 | 0.8940 | ### Framework versions - Transformers 4.28.1 - Pytorch 2.0.0+cu118 - Datasets 2.11.0 - Tokenizers 0.13.3
1,764
[ [ -0.0287017822265625, -0.046417236328125, 0.01267242431640625, 0.0171051025390625, -0.0265350341796875, -0.00826263427734375, -0.00733184814453125, -0.0083465576171875, 0.013763427734375, 0.0149688720703125, -0.04791259765625, -0.04132080078125, -0.06130981445312...
lugrenl/bert-base-banking77-pt2
2023-04-12T21:44:58.000Z
[ "transformers", "pytorch", "tensorboard", "bert", "text-classification", "generated_from_trainer", "dataset:banking77", "license:apache-2.0", "model-index", "endpoints_compatible", "region:us" ]
text-classification
lugrenl
null
null
lugrenl/bert-base-banking77-pt2
0
2
transformers
2023-04-12T20:47:18
--- license: apache-2.0 tags: - generated_from_trainer datasets: - banking77 metrics: - f1 model-index: - name: bert-base-banking77-pt2 results: - task: name: Text Classification type: text-classification dataset: name: banking77 type: banking77 config: default split: test args: default metrics: - name: F1 type: f1 value: 0.9360461829994651 --- <!-- 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. --> # bert-base-banking77-pt2 This model is a fine-tuned version of [bert-base-uncased](https://huggingface.co/bert-base-uncased) on the banking77 dataset. It achieves the following results on the evaluation set: - Loss: 0.3261 - F1: 0.9360 ## 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: 4 - eval_batch_size: 2 - seed: 42 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 - lr_scheduler_type: linear - num_epochs: 3 ### Training results | Training Loss | Epoch | Step | Validation Loss | F1 | |:-------------:|:-----:|:----:|:---------------:|:------:| | 0.5369 | 1.0 | 2501 | 0.4475 | 0.8808 | | 0.2189 | 2.0 | 5002 | 0.3341 | 0.9290 | | 0.1552 | 3.0 | 7503 | 0.3261 | 0.9360 | ### Framework versions - Transformers 4.27.1 - Pytorch 2.0.0+cu118 - Datasets 2.11.0 - Tokenizers 0.11.0
1,728
[ [ -0.0299530029296875, -0.0396728515625, 0.0121002197265625, 0.0131072998046875, -0.042266845703125, -0.0274658203125, -0.00896453857421875, -0.018157958984375, -0.00417327880859375, 0.041259765625, -0.042999267578125, -0.04339599609375, -0.052093505859375, -0...
DrewG/Tale_2_Cities
2023-04-12T21:26:27.000Z
[ "transformers", "pytorch", "tensorboard", "gpt2", "text-generation", "generated_from_trainer", "license:mit", "endpoints_compatible", "text-generation-inference", "region:us" ]
text-generation
DrewG
null
null
DrewG/Tale_2_Cities
0
2
transformers
2023-04-12T20:55:39
--- license: mit tags: - generated_from_trainer model-index: - name: Tale_2_Cities 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. --> # Tale_2_Cities This model is a fine-tuned version of [gpt2](https://huggingface.co/gpt2) on an unknown 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: 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: 30 ### Training results ### Framework versions - Transformers 4.27.4 - Pytorch 2.0.0+cu118 - Datasets 2.11.0 - Tokenizers 0.13.3
989
[ [ -0.027740478515625, -0.050323486328125, 0.037078857421875, 0.00719451904296875, -0.024627685546875, -0.0212249755859375, 0.004329681396484375, -0.031951904296875, -0.0005044937133789062, 0.0238037109375, -0.04718017578125, -0.035736083984375, -0.047332763671875,...
platzi/platzi-distilroberta-base-mrpc-glue-oscar-moreno
2023-04-12T22:13:55.000Z
[ "transformers", "pytorch", "tensorboard", "roberta", "text-classification", "generated_from_trainer", "dataset:glue", "license:apache-2.0", "model-index", "endpoints_compatible", "region:us" ]
text-classification
platzi
null
null
platzi/platzi-distilroberta-base-mrpc-glue-oscar-moreno
0
2
transformers
2023-04-12T21:47:10
--- license: apache-2.0 tags: - text-classification - generated_from_trainer datasets: - glue metrics: - accuracy - f1 model-index: - name: platzi-distilroberta-base-mrpc-glue-oscar-moreno results: - task: name: Text Classification type: text-classification dataset: name: glue type: glue config: mrpc split: validation args: mrpc metrics: - name: Accuracy type: accuracy value: 0.8161764705882353 - name: F1 type: f1 value: 0.8695652173913044 --- <!-- 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. --> # platzi-distilroberta-base-mrpc-glue-oscar-moreno This model is a fine-tuned version of [distilroberta-base](https://huggingface.co/distilroberta-base) on the glue and the mrpc datasets. It achieves the following results on the evaluation set: - Loss: 0.5426 - Accuracy: 0.8162 - F1: 0.8696 ## 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: 3 ### Training results | Training Loss | Epoch | Step | Validation Loss | Accuracy | F1 | |:-------------:|:-----:|:----:|:---------------:|:--------:|:------:| | 0.5255 | 1.09 | 500 | 0.5426 | 0.8162 | 0.8696 | | 0.3669 | 2.18 | 1000 | 0.5466 | 0.8480 | 0.8869 | ### Framework versions - Transformers 4.27.4 - Pytorch 2.0.0+cu118 - Datasets 2.11.0 - Tokenizers 0.13.3
1,884
[ [ -0.0305328369140625, -0.03570556640625, 0.006641387939453125, 0.0182952880859375, -0.030426025390625, -0.0249786376953125, -0.005229949951171875, -0.00688934326171875, 0.01181793212890625, 0.007488250732421875, -0.049652099609375, -0.038360595703125, -0.05941772...
auditi41/wav2vec2-large-xlsr-53-Bangla
2023-04-13T18:52:35.000Z
[ "transformers", "pytorch", "tensorboard", "wav2vec2", "automatic-speech-recognition", "generated_from_trainer", "dataset:common_voice_11_0", "license:apache-2.0", "model-index", "endpoints_compatible", "region:us" ]
automatic-speech-recognition
auditi41
null
null
auditi41/wav2vec2-large-xlsr-53-Bangla
1
2
transformers
2023-04-12T23:42:56
--- license: apache-2.0 tags: - generated_from_trainer datasets: - common_voice_11_0 metrics: - wer model-index: - name: wav2vec2-large-xlsr-53-Bangla results: - task: name: Automatic Speech Recognition type: automatic-speech-recognition dataset: name: common_voice_11_0 type: common_voice_11_0 config: bn split: train+validation args: bn metrics: - name: Wer type: wer value: 0.5442110214000156 --- <!-- 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. --> # wav2vec2-large-xlsr-53-Bangla This model is a fine-tuned version of [facebook/wav2vec2-large-xlsr-53](https://huggingface.co/facebook/wav2vec2-large-xlsr-53) on the common_voice_11_0 dataset. It achieves the following results on the evaluation set: - Loss: 0.6125 - Wer: 0.5442 ## 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: 0.0004 - train_batch_size: 16 - eval_batch_size: 8 - seed: 42 - gradient_accumulation_steps: 2 - total_train_batch_size: 32 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 - lr_scheduler_type: linear - lr_scheduler_warmup_steps: 500 - num_epochs: 30 ### Training results | Training Loss | Epoch | Step | Validation Loss | Wer | |:-------------:|:-----:|:----:|:---------------:|:------:| | 4.6881 | 2.28 | 600 | 1.0325 | 0.9634 | | 0.8087 | 4.56 | 1200 | 0.6090 | 0.7430 | | 0.5089 | 6.84 | 1800 | 0.5156 | 0.6615 | | 0.3864 | 9.13 | 2400 | 0.5287 | 0.6676 | | 0.3064 | 11.41 | 3000 | 0.5411 | 0.6278 | | 0.2535 | 13.69 | 3600 | 0.5206 | 0.6149 | | 0.216 | 15.97 | 4200 | 0.5596 | 0.6120 | | 0.1852 | 18.25 | 4800 | 0.5658 | 0.5821 | | 0.1653 | 20.53 | 5400 | 0.5938 | 0.5521 | | 0.1499 | 22.81 | 6000 | 0.5825 | 0.5645 | | 0.1323 | 25.09 | 6600 | 0.6151 | 0.5593 | | 0.122 | 27.38 | 7200 | 0.6046 | 0.5556 | | 0.1118 | 29.66 | 7800 | 0.6125 | 0.5442 | ### Framework versions - Transformers 4.24.0 - Pytorch 2.0.0+cu118 - Datasets 2.11.0 - Tokenizers 0.13.3
2,522
[ [ -0.033905029296875, -0.0330810546875, -0.002201080322265625, 0.01416015625, -0.017242431640625, -0.0129547119140625, -0.01476287841796875, -0.0205078125, 0.0129547119140625, 0.02532958984375, -0.059844970703125, -0.04022216796875, -0.0487060546875, -0.019393...
Muhsabrys/autotrain-iuexistmulti-49035118635
2023-04-13T02:15:21.000Z
[ "transformers", "pytorch", "bert", "text-classification", "autotrain", "unk", "dataset:Muhsabrys/autotrain-data-iuexistmulti", "co2_eq_emissions", "endpoints_compatible", "region:us" ]
text-classification
Muhsabrys
null
null
Muhsabrys/autotrain-iuexistmulti-49035118635
0
2
transformers
2023-04-13T02:13:14
--- tags: - autotrain - text-classification language: - unk widget: - text: "I love AutoTrain 🤗" datasets: - Muhsabrys/autotrain-data-iuexistmulti co2_eq_emissions: emissions: 0.8019084818135189 --- # Model Trained Using AutoTrain - Problem type: Multi-class Classification - Model ID: 49035118635 - CO2 Emissions (in grams): 0.8019 ## Validation Metrics - Loss: 0.691 - Accuracy: 0.743 - Macro F1: 0.521 - Micro F1: 0.743 - Weighted F1: 0.704 - Macro Precision: 0.495 - Micro Precision: 0.743 - Weighted Precision: 0.668 - Macro Recall: 0.550 - Micro Recall: 0.743 - Weighted Recall: 0.743 ## 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/Muhsabrys/autotrain-iuexistmulti-49035118635 ``` Or Python API: ``` from transformers import AutoModelForSequenceClassification, AutoTokenizer model = AutoModelForSequenceClassification.from_pretrained("Muhsabrys/autotrain-iuexistmulti-49035118635", use_auth_token=True) tokenizer = AutoTokenizer.from_pretrained("Muhsabrys/autotrain-iuexistmulti-49035118635", use_auth_token=True) inputs = tokenizer("I love AutoTrain", return_tensors="pt") outputs = model(**inputs) ```
1,301
[ [ -0.034423828125, -0.021148681640625, 0.0106048583984375, 0.0106201171875, -0.00018155574798583984, 0.005218505859375, 0.0005030632019042969, -0.01424407958984375, -0.00426483154296875, 0.0034732818603515625, -0.048736572265625, -0.0305938720703125, -0.0572814941...
vishwapatel123/Toxic-comment
2023-04-13T02:28:57.000Z
[ "transformers", "pytorch", "bert", "text-classification", "en", "license:afl-3.0", "endpoints_compatible", "region:us" ]
text-classification
vishwapatel123
null
null
vishwapatel123/Toxic-comment
0
2
transformers
2023-04-13T02:25:01
--- license: afl-3.0 language: - en metrics: - accuracy library_name: transformers pipeline_tag: text-classification --- ## Name Vishwa Patel ## Project Toxic Comment Classification ## Model description This model is a fine-tuned version of the [bert-base-uncased](https://huggingface.co/transformers/model_doc/bert.html) model to classify toxic comments. ## Training data The training data comes from this [Kaggle competition](https://www.kaggle.com/c/jigsaw-unintended-bias-in-toxicity-classification/data). We use 90% of the `train.csv` data to train the model.
570
[ [ -0.019775390625, -0.0303802490234375, 0.0241241455078125, 0.006725311279296875, -0.020294189453125, 0.01276397705078125, 0.0089874267578125, -0.0190887451171875, 0.0003657341003417969, 0.043701171875, -0.044403076171875, -0.031402587890625, -0.05303955078125, ...
Muhsabrys/autotrain-iuexist_twhin-49038118649
2023-04-13T02:35:17.000Z
[ "transformers", "pytorch", "bert", "text-classification", "autotrain", "unk", "dataset:Muhsabrys/autotrain-data-iuexist_twhin", "co2_eq_emissions", "endpoints_compatible", "region:us" ]
text-classification
Muhsabrys
null
null
Muhsabrys/autotrain-iuexist_twhin-49038118649
0
2
transformers
2023-04-13T02:31:37
--- tags: - autotrain - text-classification language: - unk widget: - text: "I love AutoTrain 🤗" datasets: - Muhsabrys/autotrain-data-iuexist_twhin co2_eq_emissions: emissions: 1.410217361850194 --- # Model Trained Using AutoTrain - Problem type: Multi-class Classification - Model ID: 49038118649 - CO2 Emissions (in grams): 1.4102 ## Validation Metrics - Loss: 0.636 - Accuracy: 0.766 - Macro F1: 0.537 - Micro F1: 0.766 - Weighted F1: 0.725 - Macro Precision: 0.511 - Micro Precision: 0.766 - Weighted Precision: 0.690 - Macro Recall: 0.567 - Micro Recall: 0.766 - Weighted Recall: 0.766 ## 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/Muhsabrys/autotrain-iuexist_twhin-49038118649 ``` Or Python API: ``` from transformers import AutoModelForSequenceClassification, AutoTokenizer model = AutoModelForSequenceClassification.from_pretrained("Muhsabrys/autotrain-iuexist_twhin-49038118649", use_auth_token=True) tokenizer = AutoTokenizer.from_pretrained("Muhsabrys/autotrain-iuexist_twhin-49038118649", use_auth_token=True) inputs = tokenizer("I love AutoTrain", return_tensors="pt") outputs = model(**inputs) ```
1,304
[ [ -0.031341552734375, -0.0219879150390625, 0.0113525390625, 0.009918212890625, -0.0011653900146484375, 0.00787353515625, -0.0009832382202148438, -0.01470947265625, -0.007083892822265625, 0.0036182403564453125, -0.049102783203125, -0.033294677734375, -0.05691528320...