initial

README.md

@@ -0,0 +1,84 @@
+---
+language:
+- en
+tags:
+- generated_from_trainer
+datasets:
+- mnli
+metrics:
+- accuracy
+model-index:
+- name: glue
+  results:
+  - task:
+      name: Text Classification
+      type: text-classification
+    dataset:
+      name: GLUE MNLI
+      type: glue
+      args: mnli
+    metrics:
+    - name: Accuracy
+      type: accuracy
+      value: 0.834519934906428
+---
+
+<!-- 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. -->
+
+# mnli
+
+This model is a fine-tuned version of [allenai/scibert_scivocab_uncased](https://huggingface.co/allenai/scibert_scivocab_uncased) on the GLUE MNLI dataset.
+It achieves the following results on the evaluation set:
+- Loss: 0.4917
+- Accuracy: 0.8345
+
+## Model description
+
+This is the pretrained model presented in [SciBERT: A Pretrained Language Model for Scientific Text](https://www.aclweb.org/anthology/D19-1371/), which is a BERT model trained on scientific text, then finetuned on GLUE MNLI for zero-shot classification.
+
+The training corpus was papers taken from [Semantic Scholar](https://www.semanticscholar.org). Corpus size is 1.14M papers, 3.1B tokens. We use the full text of the papers in training, not just abstracts.
+
+SciBERT has its own wordpiece vocabulary (scivocab) that's built to best match the training corpus.
+
+## Intended uses & limitations
+
+Zero-shot classification of scientific texts. Note that this model is outperformed by multiple models and was uploaded for research purposes. For actually classifying scientific text, I recommend looking into [Deberta v3 Large tuned on MNLI](https://huggingface.co/navteca/nli-deberta-v3-large) which according to my benchmark on abstracts performs best at current date (7/10/22).
+
+## Training and evaluation data
+
+GLUE MNLI
+
+## Training procedure
+
+### Training hyperparameters
+
+The following hyperparameters were used during training:
+- learning_rate: 2e-05
+- train_batch_size: 32
+- 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
+- mixed_precision_training: Native AMP
+
+### Framework versions
+
+- Transformers 4.22.2
+- Pytorch 1.11.0+cu113
+- Datasets 2.5.1
+- Tokenizers 0.12.1
+
+
+If using these models, please cite the following paper:
+```
+@inproceedings{beltagy-etal-2019-scibert,
+    title = "SciBERT: A Pretrained Language Model for Scientific Text",
+    author = "Beltagy, Iz  and Lo, Kyle  and Cohan, Arman",
+    booktitle = "EMNLP",
+    year = "2019",
+    publisher = "Association for Computational Linguistics",
+    url = "https://www.aclweb.org/anthology/D19-1371"
+}
+```

all_results.json

@@ -0,0 +1,22 @@
+{
+    "epoch": 3.0,
+    "epoch_mm": 3.0,
+    "eval_accuracy": 0.823841059602649,
+    "eval_accuracy_mm": 0.834519934906428,
+    "eval_loss": 0.5203812718391418,
+    "eval_loss_mm": 0.4916685223579407,
+    "eval_runtime": 17.8999,
+    "eval_runtime_mm": 20.4415,
+    "eval_samples": 9815,
+    "eval_samples_mm": 9832,
+    "eval_samples_per_second": 548.327,
+    "eval_samples_per_second_mm": 480.983,
+    "eval_steps_per_second": 68.548,
+    "eval_steps_per_second_mm": 60.123,
+    "total_flos": 7.74938740264658e+16,
+    "train_loss": 0.42257067142803223,
+    "train_runtime": 4332.4547,
+    "train_samples": 392702,
+    "train_samples_per_second": 271.926,
+    "train_steps_per_second": 8.498
+}

config.json

@@ -0,0 +1,37 @@
+{
+  "_name_or_path": "allenai/scibert_scivocab_uncased",
+  "architectures": [
+    "BertForSequenceClassification"
+  ],
+  "attention_probs_dropout_prob": 0.1,
+  "classifier_dropout": null,
+  "finetuning_task": "mnli",
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 768,
+  "id2label": {
+    "0": "entailment",
+    "1": "neutral",
+    "2": "contradiction"
+  },
+  "initializer_range": 0.02,
+  "intermediate_size": 3072,
+  "label2id": {
+    "contradiction": 2,
+    "entailment": 0,
+    "neutral": 1
+  },
+  "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",
+  "problem_type": "single_label_classification",
+  "torch_dtype": "float32",
+  "transformers_version": "4.22.2",
+  "type_vocab_size": 2,
+  "use_cache": true,
+  "vocab_size": 31090
+}

eval_results.json

@@ -0,0 +1,16 @@
+{
+    "epoch": 3.0,
+    "epoch_mm": 3.0,
+    "eval_accuracy": 0.823841059602649,
+    "eval_accuracy_mm": 0.834519934906428,
+    "eval_loss": 0.5203812718391418,
+    "eval_loss_mm": 0.4916685223579407,
+    "eval_runtime": 17.8999,
+    "eval_runtime_mm": 20.4415,
+    "eval_samples": 9815,
+    "eval_samples_mm": 9832,
+    "eval_samples_per_second": 548.327,
+    "eval_samples_per_second_mm": 480.983,
+    "eval_steps_per_second": 68.548,
+    "eval_steps_per_second_mm": 60.123
+}

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1e640c10f6e574204b0cfd703c8d80805269a934b6ddba02321e7096b33b17b8
+size 439754093

special_tokens_map.json

@@ -0,0 +1,7 @@
+{
+  "cls_token": "[CLS]",
+  "mask_token": "[MASK]",
+  "pad_token": "[PAD]",
+  "sep_token": "[SEP]",
+  "unk_token": "[UNK]"
+}

tokenizer_config.json

@@ -0,0 +1,16 @@
+{
+  "cls_token": "[CLS]",
+  "do_basic_tokenize": true,
+  "do_lower_case": true,
+  "mask_token": "[MASK]",
+  "name_or_path": "allenai/scibert_scivocab_uncased",
+  "never_split": null,
+  "pad_token": "[PAD]",
+  "sep_token": "[SEP]",
+  "model_max_length": 512,
+  "special_tokens_map_file": null,
+  "strip_accents": null,
+  "tokenize_chinese_chars": true,
+  "tokenizer_class": "BertTokenizer",
+  "unk_token": "[UNK]"
+}

train_results.json

@@ -0,0 +1,9 @@
+{
+    "epoch": 3.0,
+    "total_flos": 7.74938740264658e+16,
+    "train_loss": 0.42257067142803223,
+    "train_runtime": 4332.4547,
+    "train_samples": 392702,
+    "train_samples_per_second": 271.926,
+    "train_steps_per_second": 8.498
+}

trainer_state.json

@@ -0,0 +1,463 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.0,
+  "global_step": 36816,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9728378965667103e-05,
+      "loss": 0.8341,
+      "step": 500
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9456757931334205e-05,
+      "loss": 0.6968,
+      "step": 1000
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9185136897001307e-05,
+      "loss": 0.6557,
+      "step": 1500
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8913515862668405e-05,
+      "loss": 0.6154,
+      "step": 2000
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8642438070404173e-05,
+      "loss": 0.6167,
+      "step": 2500
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8370817036071274e-05,
+      "loss": 0.6043,
+      "step": 3000
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8099196001738376e-05,
+      "loss": 0.5908,
+      "step": 3500
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7828118209474144e-05,
+      "loss": 0.5746,
+      "step": 4000
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7556497175141245e-05,
+      "loss": 0.5532,
+      "step": 4500
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7284876140808347e-05,
+      "loss": 0.5657,
+      "step": 5000
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7013255106475445e-05,
+      "loss": 0.5564,
+      "step": 5500
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6741634072142547e-05,
+      "loss": 0.5406,
+      "step": 6000
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.647001303780965e-05,
+      "loss": 0.5518,
+      "step": 6500
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.619839200347675e-05,
+      "loss": 0.546,
+      "step": 7000
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5927314211212518e-05,
+      "loss": 0.5373,
+      "step": 7500
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.565569317687962e-05,
+      "loss": 0.5262,
+      "step": 8000
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.538407214254672e-05,
+      "loss": 0.5285,
+      "step": 8500
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5112451108213821e-05,
+      "loss": 0.5331,
+      "step": 9000
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4840830073880923e-05,
+      "loss": 0.5171,
+      "step": 9500
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4569209039548025e-05,
+      "loss": 0.5122,
+      "step": 10000
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4297588005215127e-05,
+      "loss": 0.5187,
+      "step": 10500
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4025966970882225e-05,
+      "loss": 0.5141,
+      "step": 11000
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3754889178617994e-05,
+      "loss": 0.5068,
+      "step": 11500
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3483268144285096e-05,
+      "loss": 0.5089,
+      "step": 12000
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3211647109952194e-05,
+      "loss": 0.4605,
+      "step": 12500
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2940026075619296e-05,
+      "loss": 0.4191,
+      "step": 13000
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2668948283355065e-05,
+      "loss": 0.4084,
+      "step": 13500
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2397327249022167e-05,
+      "loss": 0.4139,
+      "step": 14000
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2126249456757933e-05,
+      "loss": 0.3967,
+      "step": 14500
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1854628422425034e-05,
+      "loss": 0.4078,
+      "step": 15000
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1583007388092136e-05,
+      "loss": 0.4031,
+      "step": 15500
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1311386353759234e-05,
+      "loss": 0.4051,
+      "step": 16000
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1039765319426336e-05,
+      "loss": 0.4085,
+      "step": 16500
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0768144285093438e-05,
+      "loss": 0.4107,
+      "step": 17000
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.049652325076054e-05,
+      "loss": 0.3993,
+      "step": 17500
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0224902216427641e-05,
+      "loss": 0.4082,
+      "step": 18000
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.953824424163409e-06,
+      "loss": 0.3987,
+      "step": 18500
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.682203389830509e-06,
+      "loss": 0.4062,
+      "step": 19000
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.41058235549761e-06,
+      "loss": 0.4048,
+      "step": 19500
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.138961321164712e-06,
+      "loss": 0.408,
+      "step": 20000
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.867883528900478e-06,
+      "loss": 0.4056,
+      "step": 20500
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.59626249456758e-06,
+      "loss": 0.3995,
+      "step": 21000
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.325184702303347e-06,
+      "loss": 0.4005,
+      "step": 21500
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.053563667970449e-06,
+      "loss": 0.3974,
+      "step": 22000
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.781942633637549e-06,
+      "loss": 0.3992,
+      "step": 22500
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.510321599304651e-06,
+      "loss": 0.405,
+      "step": 23000
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.238700564971752e-06,
+      "loss": 0.3901,
+      "step": 23500
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.967079530638852e-06,
+      "loss": 0.3989,
+      "step": 24000
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.695458496305954e-06,
+      "loss": 0.4007,
+      "step": 24500
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.423837461973056e-06,
+      "loss": 0.301,
+      "step": 25000
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.152759669708823e-06,
+      "loss": 0.3016,
+      "step": 25500
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.881138635375923e-06,
+      "loss": 0.3001,
+      "step": 26000
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.609517601043025e-06,
+      "loss": 0.2919,
+      "step": 26500
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.337896566710127e-06,
+      "loss": 0.2992,
+      "step": 27000
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.066818774445893e-06,
+      "loss": 0.2961,
+      "step": 27500
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.79574098218166e-06,
+      "loss": 0.3074,
+      "step": 28000
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.524119947848762e-06,
+      "loss": 0.2926,
+      "step": 28500
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.252498913515863e-06,
+      "loss": 0.2947,
+      "step": 29000
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.980877879182964e-06,
+      "loss": 0.299,
+      "step": 29500
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7092568448500653e-06,
+      "loss": 0.3009,
+      "step": 30000
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.437635810517167e-06,
+      "loss": 0.3013,
+      "step": 30500
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1665580182529336e-06,
+      "loss": 0.2908,
+      "step": 31000
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.894936983920035e-06,
+      "loss": 0.2921,
+      "step": 31500
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.6233159495871362e-06,
+      "loss": 0.295,
+      "step": 32000
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3516949152542375e-06,
+      "loss": 0.2833,
+      "step": 32500
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.080073880921339e-06,
+      "loss": 0.3009,
+      "step": 33000
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.80845284658844e-06,
+      "loss": 0.2795,
+      "step": 33500
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5368318122555412e-06,
+      "loss": 0.2861,
+      "step": 34000
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 1.2652107779226425e-06,
+      "loss": 0.2991,
+      "step": 34500
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 9.941329856584096e-07,
+      "loss": 0.2836,
+      "step": 35000
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 7.225119513255108e-07,
+      "loss": 0.2925,
+      "step": 35500
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 4.5089091699261194e-07,
+      "loss": 0.2927,
+      "step": 36000
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 1.7926988265971318e-07,
+      "loss": 0.2921,
+      "step": 36500
+    },
+    {
+      "epoch": 3.0,
+      "step": 36816,
+      "total_flos": 7.74938740264658e+16,
+      "train_loss": 0.42257067142803223,
+      "train_runtime": 4332.4547,
+      "train_samples_per_second": 271.926,
+      "train_steps_per_second": 8.498
+    }
+  ],
+  "max_steps": 36816,
+  "num_train_epochs": 3,
+  "total_flos": 7.74938740264658e+16,
+  "trial_name": null,
+  "trial_params": null
+}

training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5f118d321b6ab58292c298c3ea2c75a3211f3fb0b13bb9da8c24b3fa647f22be
+size 3439