Initial Commit

README.md

@@ -0,0 +1,80 @@
+---
+license: mit
+tags:
+- generated_from_trainer
+datasets:
+- pritamdeka/cord-19-abstract
+metrics:
+- accuracy
+model-index:
+- name: pubmedbert-abstract-cord19
+  results:
+  - task:
+      name: Masked Language Modeling
+      type: fill-mask
+    dataset:
+      name: pritamdeka/cord-19-abstract
+      type: pritamdeka/cord-19-abstract
+      args: fulltext
+    metrics:
+    - name: Accuracy
+      type: accuracy
+      value: 0.7246798699728464
+---
+
+<!-- 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. -->
+
+# pubmedbert-abstract-cord19
+
+This model is a fine-tuned version of [microsoft/BiomedNLP-PubMedBERT-base-uncased-abstract-fulltext](https://huggingface.co/microsoft/BiomedNLP-PubMedBERT-base-uncased-abstract-fulltext) on the pritamdeka/cord-19-abstract dataset.
+It achieves the following results on the evaluation set:
+- Loss: 1.2371
+- Accuracy: 0.7247
+
+## 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=(0.9,0.95) and epsilon=1e-08
+- lr_scheduler_type: linear
+- lr_scheduler_warmup_steps: 10000
+- num_epochs: 4.0
+- mixed_precision_training: Native AMP
+
+### Training results
+
+| Training Loss | Epoch | Step  | Validation Loss | Accuracy |
+|:-------------:|:-----:|:-----:|:---------------:|:--------:|
+| 1.27          | 0.53  | 5000  | 1.2425          | 0.7236   |
+| 1.2634        | 1.06  | 10000 | 1.3123          | 0.7141   |
+| 1.3041        | 1.59  | 15000 | 1.3583          | 0.7072   |
+| 1.3829        | 2.12  | 20000 | 1.3590          | 0.7121   |
+| 1.3069        | 2.65  | 25000 | 1.3506          | 0.7154   |
+| 1.2921        | 3.18  | 30000 | 1.3448          | 0.7160   |
+| 1.2731        | 3.7   | 35000 | 1.3375          | 0.7178   |
+
+
+### Framework versions
+
+- Transformers 4.17.0.dev0
+- Pytorch 1.10.0+cu111
+- Datasets 1.18.3
+- Tokenizers 0.11.0

all_results.json

@@ -0,0 +1,15 @@
+{
+    "epoch": 4.0,
+    "eval_accuracy": 0.7246798699728464,
+    "eval_loss": 1.2370661497116089,
+    "eval_runtime": 183.5428,
+    "eval_samples": 8110,
+    "eval_samples_per_second": 44.186,
+    "eval_steps_per_second": 2.762,
+    "perplexity": 3.445490069726643,
+    "train_loss": 1.2965481223991413,
+    "train_runtime": 37723.3327,
+    "train_samples": 151146,
+    "train_samples_per_second": 16.027,
+    "train_steps_per_second": 1.002
+}

config.json

@@ -0,0 +1,25 @@
+{
+  "_name_or_path": "microsoft/BiomedNLP-PubMedBERT-base-uncased-abstract-fulltext",
+  "architectures": [
+    "BertForMaskedLM"
+  ],
+  "attention_probs_dropout_prob": 0.1,
+  "classifier_dropout": null,
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 768,
+  "initializer_range": 0.02,
+  "intermediate_size": 3072,
+  "layer_norm_eps": 1e-12,
+  "max_position_embeddings": 512,
+  "model_type": "bert",
+  "num_attention_heads": 12,
+  "num_hidden_layers": 12,
+  "pad_token_id": 0,
+  "position_embedding_type": "absolute",
+  "torch_dtype": "float32",
+  "transformers_version": "4.17.0.dev0",
+  "type_vocab_size": 2,
+  "use_cache": true,
+  "vocab_size": 30522
+}

eval_results.json

@@ -0,0 +1,10 @@
+{
+    "epoch": 4.0,
+    "eval_accuracy": 0.7246798699728464,
+    "eval_loss": 1.2370661497116089,
+    "eval_runtime": 183.5428,
+    "eval_samples": 8110,
+    "eval_samples_per_second": 44.186,
+    "eval_steps_per_second": 2.762,
+    "perplexity": 3.445490069726643
+}

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cf6a91f5d74dc8bc3e66db684ecb445897af04b9ebe460eae4fc36436e3e5ae3
+size 438141995

special_tokens_map.json

@@ -0,0 +1 @@
+{"unk_token": "[UNK]", "sep_token": "[SEP]", "pad_token": "[PAD]", "cls_token": "[CLS]", "mask_token": "[MASK]"}

tokenizer_config.json

@@ -0,0 +1 @@
+{"do_lower_case": true, "unk_token": "[UNK]", "sep_token": "[SEP]", "pad_token": "[PAD]", "cls_token": "[CLS]", "mask_token": "[MASK]", "tokenize_chinese_chars": true, "strip_accents": null, "special_tokens_map_file": null, "name_or_path": "microsoft/BiomedNLP-PubMedBERT-base-uncased-abstract-fulltext", "do_basic_tokenize": true, "never_split": null, "tokenizer_class": "BertTokenizer"}

train_results.json

@@ -0,0 +1,8 @@
+{
+    "epoch": 4.0,
+    "train_loss": 1.2965481223991413,
+    "train_runtime": 37723.3327,
+    "train_samples": 151146,
+    "train_samples_per_second": 16.027,
+    "train_steps_per_second": 1.002
+}

trainer_state.json

@@ -0,0 +1,130 @@
+{
+  "best_metric": 1.2424818277359009,
+  "best_model_checkpoint": "/content/pubmedbert-abstract-cord19/checkpoint-5000",
+  "epoch": 4.0,
+  "global_step": 37788,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.53,
+      "learning_rate": 2.4970000000000003e-05,
+      "loss": 1.27,
+      "step": 5000
+    },
+    {
+      "epoch": 0.53,
+      "eval_accuracy": 0.7236279677533664,
+      "eval_loss": 1.2424818277359009,
+      "eval_runtime": 183.46,
+      "eval_samples_per_second": 44.206,
+      "eval_steps_per_second": 2.764,
+      "step": 5000
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 4.9965e-05,
+      "loss": 1.2634,
+      "step": 10000
+    },
+    {
+      "epoch": 1.06,
+      "eval_accuracy": 0.7140570760001546,
+      "eval_loss": 1.312286138534546,
+      "eval_runtime": 183.3534,
+      "eval_samples_per_second": 44.232,
+      "eval_steps_per_second": 2.765,
+      "step": 10000
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.1024902835756445e-05,
+      "loss": 1.3041,
+      "step": 15000
+    },
+    {
+      "epoch": 1.59,
+      "eval_accuracy": 0.7072068247593172,
+      "eval_loss": 1.3583375215530396,
+      "eval_runtime": 183.3479,
+      "eval_samples_per_second": 44.233,
+      "eval_steps_per_second": 2.765,
+      "step": 15000
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 3.2033611630919824e-05,
+      "loss": 1.3829,
+      "step": 20000
+    },
+    {
+      "epoch": 2.12,
+      "eval_accuracy": 0.7120832508972437,
+      "eval_loss": 1.3589657545089722,
+      "eval_runtime": 183.3751,
+      "eval_samples_per_second": 44.226,
+      "eval_steps_per_second": 2.765,
+      "step": 20000
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.303692241255218e-05,
+      "loss": 1.3069,
+      "step": 25000
+    },
+    {
+      "epoch": 2.65,
+      "eval_accuracy": 0.7153998687867195,
+      "eval_loss": 1.3505500555038452,
+      "eval_runtime": 183.3915,
+      "eval_samples_per_second": 44.222,
+      "eval_steps_per_second": 2.765,
+      "step": 25000
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 1.4040233194184541e-05,
+      "loss": 1.2921,
+      "step": 30000
+    },
+    {
+      "epoch": 3.18,
+      "eval_accuracy": 0.7159658967235916,
+      "eval_loss": 1.3447710275650024,
+      "eval_runtime": 183.3443,
+      "eval_samples_per_second": 44.234,
+      "eval_steps_per_second": 2.765,
+      "step": 30000
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 5.0471426515042465e-06,
+      "loss": 1.2731,
+      "step": 35000
+    },
+    {
+      "epoch": 3.7,
+      "eval_accuracy": 0.7177905274023574,
+      "eval_loss": 1.3374903202056885,
+      "eval_runtime": 183.3182,
+      "eval_samples_per_second": 44.24,
+      "eval_steps_per_second": 2.766,
+      "step": 35000
+    },
+    {
+      "epoch": 4.0,
+      "step": 37788,
+      "total_flos": 1.3985984051532e+17,
+      "train_loss": 1.2965481223991413,
+      "train_runtime": 37723.3327,
+      "train_samples_per_second": 16.027,
+      "train_steps_per_second": 1.002
+    }
+  ],
+  "max_steps": 37788,
+  "num_train_epochs": 4,
+  "total_flos": 1.3985984051532e+17,
+  "trial_name": null,
+  "trial_params": null
+}

training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b3fcee15797b8744d6ddeafec41ef0bee56e60b9c2db01a49d7d382a578693e5
+size 3055