Merge branch 'glue-example' into tf2

.circleci/config.yml

@@ -81,6 +81,7 @@ jobs:
             - checkout
             - run: sudo pip install --progress-bar off -r docs/requirements.txt
             - run: sudo pip install --progress-bar off -r requirements.txt
+            - run: cd docs/source && ln -s ../../examples/README.md examples.md && cd -
             - run: cd docs && make clean && make html && scp -r -oStrictHostKeyChecking=no _build/html/* $doc:$dir
 workflow_filters: &workflow_filters
     filters:

.gitignore

@@ -130,4 +130,5 @@ runs
 examples/runs
 
 # data
-data
\ No newline at end of file
+/data
+serialization_dir
\ No newline at end of file

README.md

@@ -21,6 +21,7 @@ These implementations have been tested on several datasets (see the example scri
 | Section | Description |
 |-|-|
 | [Installation](#installation) | How to install the package |
+| [Online demo](#online-demo) | Experimenting with this repo’s text generation capabilities |
 | [Quick tour: Usage](#quick-tour) | Tokenizers & models usage: Bert and GPT-2 |
 | [Quick tour: Fine-tuning/usage scripts](#quick-tour-of-the-fine-tuningusage-scripts) | Using provided scripts: GLUE, SQuAD and Text generation |
 | [Migrating from pytorch-pretrained-bert to pytorch-transformers](#Migrating-from-pytorch-pretrained-bert-to-pytorch-transformers) | Migrating your code from pytorch-pretrained-bert to pytorch-transformers |
@@ -68,6 +69,14 @@ It contains an example of a conversion script from a Pytorch trained Transformer
 At some point in the future, you'll be able to seamlessly move from pre-training or fine-tuning models in PyTorch to productizing them in CoreML,
 or prototype a model or an app in CoreML then research its hyperparameters or architecture from PyTorch. Super exciting!
 
+## Online demo
+
+**[Write With Transformer](https://transformer.huggingface.co)**, built by the Hugging Face team at transformer.huggingface.co, is the official demo of this repo’s text generation capabilities.
+You can use it to experiment with completions generated by `GPT2Model`, `TransfoXLModel`, and `XLNetModel`.
+
+> “🦄 Write with transformer is to writing what calculators are to calculus.”
+
+![write_with_transformer](https://transformer.huggingface.co/front/assets/thumbnail-large.png)
 
 ## Quick tour
 
@@ -279,7 +288,7 @@ This is the model provided as `bert-large-uncased-whole-word-masking-finetuned-s
 ### `run_generation.py`: Text generation with GPT, GPT-2, Transformer-XL and XLNet
 
 A conditional generation script is also included to generate text from a prompt.
-The generation script includes the [tricks](https://github.com/rusiaaman/XLNet-gen#methodology) proposed by by Aman Rusia to get high quality generation with memory models like Transformer-XL and XLNet (include a predefined text to make short inputs longer).
+The generation script includes the [tricks](https://github.com/rusiaaman/XLNet-gen#methodology) proposed by Aman Rusia to get high quality generation with memory models like Transformer-XL and XLNet (include a predefined text to make short inputs longer).
 
 Here is how to run the script with the small version of OpenAI GPT-2 model:
 

docs/README.md

@@ -34,6 +34,13 @@ pip install recommonmark
 
 ## Building the documentation
 
+Make sure that there is a symlink from the `example` file (in /examples) inside the source folder. Run the followig 
+command to generate it:
+
+```bash
+ln -s ../../examples/README.md source/examples.md
+```
+
 Once you have setup `sphinx`, you can build the documentation by running the following command in the `/docs` folder:
 
 ```bash

docs/requirements.txt

@@ -26,3 +26,4 @@ sphinxcontrib-jsmath==1.0.1
 sphinxcontrib-qthelp==1.0.2
 sphinxcontrib-serializinghtml==1.1.3
 urllib3==1.25.3
+sphinx-markdown-tables==0.0.9
\ No newline at end of file

docs/source/conf.py

@@ -43,7 +43,8 @@ extensions = [
     'sphinx.ext.coverage',
     'sphinx.ext.napoleon',
     'recommonmark',
-    'sphinx.ext.viewcode'
+    'sphinx.ext.viewcode',
+    'sphinx_markdown_tables'
 ]
 
 # Add any paths that contain templates here, relative to this directory.

docs/source/pretrained_models.rst

@@ -79,10 +79,10 @@ Here is the full list of the currently provided pretrained models together with
 |                   |                                                            | | XLM English model                                                                                                                   |
 |                   +------------------------------------------------------------+---------------------------------------------------------------------------------------------------------------------------------------+
 |                   | ``xlm-mlm-ende-1024``                                      | | 6-layer, 1024-hidden, 8-heads                                                                                                       |
-|                   |                                                            | | XLM English-German Multi-language model                                                                                             |
+|                   |                                                            | | XLM English-German model trained on the concatenation of English and German wikipedia                                               |
 |                   +------------------------------------------------------------+---------------------------------------------------------------------------------------------------------------------------------------+
 |                   | ``xlm-mlm-enfr-1024``                                      | | 6-layer, 1024-hidden, 8-heads                                                                                                       |
-|                   |                                                            | | XLM English-French Multi-language model                                                                                             |
+|                   |                                                            | | XLM English-French model trained on the concatenation of English and French wikipedia                                               |
 |                   +------------------------------------------------------------+---------------------------------------------------------------------------------------------------------------------------------------+
 |                   | ``xlm-mlm-enro-1024``                                      | | 6-layer, 1024-hidden, 8-heads                                                                                                       |
 |                   |                                                            | | XLM English-Romanian Multi-language model                                                                                           |
@@ -93,11 +93,11 @@ Here is the full list of the currently provided pretrained models together with
 |                   | ``xlm-mlm-tlm-xnli15-1024``                                | | 12-layer, 1024-hidden, 8-heads                                                                                                      |
 |                   |                                                            | | XLM Model pre-trained with MLM + TLM on the `15 XNLI languages <https://github.com/facebookresearch/XNLI>`__.                       |
 |                   +------------------------------------------------------------+---------------------------------------------------------------------------------------------------------------------------------------+
-|                   | ``xlm-clm-enfr-1024``                                      | | 12-layer, 1024-hidden, 8-heads                                                                                                      |
-|                   |                                                            | | XLM English model trained with CLM (Causal Language Modeling)                                                                       |
+|                   | ``xlm-clm-enfr-1024``                                      | | 6-layer, 1024-hidden, 8-heads                                                                                                       |
+|                   |                                                            | | XLM English-French model trained with CLM (Causal Language Modeling) on the concatenation of English and French wikipedia           |
 |                   +------------------------------------------------------------+---------------------------------------------------------------------------------------------------------------------------------------+
 |                   | ``xlm-clm-ende-1024``                                      | | 6-layer, 1024-hidden, 8-heads                                                                                                       |
-|                   |                                                            | | XLM English-German Multi-language model trained with CLM (Causal Language Modeling)                                                 |
+|                   |                                                            | | XLM English-German model trained with CLM (Causal Language Modeling) on the concatenation of English and German wikipedia           |
 +-------------------+------------------------------------------------------------+---------------------------------------------------------------------------------------------------------------------------------------+
 | RoBERTa           | ``roberta-base``                                           | | 12-layer, 768-hidden, 12-heads, 125M parameters                                                                                     |
 |                   |                                                            | | RoBERTa using the BERT-base architecture                                                                                            |

examples/README.md

+# Examples
+
+In this section a few examples are put together. All of these examples work for several models, making use of the very
+similar API between the different models.
+
+| Section                    | Description                                                                                                                                                |
+|----------------------------|------------------------------------------------------------------------------------------------------------------------------------------------------------|
+| [Language Model fine-tuning](#language-model-fine-tuning) | Fine-tuning the library models for language modeling on a text dataset. Causal language modeling for GPT/GPT-2, masked language modeling for BERT/RoBERTa. |
+| [Language Generation](#language-generation) | Conditional text generation using the auto-regressive models of the library: GPT, GPT-2, Transformer-XL and XLNet.                                         |
+| [GLUE](#glue) | Examples running BERT/XLM/XLNet/RoBERTa on the 9 GLUE tasks. Examples feature distributed training as well as half-precision.                              |
+| [SQuAD](#squad) | Using BERT for question answering, examples with distributed training.                                                                                  |
+| [Multiple Choice](#multiple choice) | Examples running BERT/XLNet/RoBERTa on the SWAG/RACE/ARC tasks. 
+
+## Language model fine-tuning
+
+Based on the script [`run_lm_finetuning.py`](https://github.com/huggingface/pytorch-transformers/blob/master/examples/run_lm_finetuning.py).
+
+Fine-tuning the library models for language modeling on a text dataset for GPT, GPT-2, BERT and RoBERTa (DistilBERT 
+to be added soon). GPT and GPT-2 are fine-tuned using a causal language modeling (CLM) loss while BERT and RoBERTa 
+are fine-tuned using a masked language modeling (MLM) loss.
+
+Before running the following example, you should get a file that contains text on which the language model will be
+fine-tuned. A good example of such text is the [WikiText-2 dataset](https://blog.einstein.ai/the-wikitext-long-term-dependency-language-modeling-dataset/).
+
+We will refer to two different files: `$TRAIN_FILE`, which contains text for training, and `$TEST_FILE`, which contains
+text that will be used for evaluation.
+
+### GPT-2/GPT and causal language modeling
+
+The following example fine-tunes GPT-2 on WikiText-2. We're using the raw WikiText-2 (no tokens were replaced before
+the tokenization). The loss here is that of causal language modeling.
+
+```bash
+export TRAIN_FILE=/path/to/dataset/wiki.train.raw
+export TEST_FILE=/path/to/dataset/wiki.test.raw
+
+python run_lm_finetuning.py \
+    --output_dir=output \
+    --model_type=gpt2 \
+    --model_name_or_path=gpt2 \
+    --do_train \
+    --train_data_file=$TRAIN_FILE \
+    --do_eval \
+    --eval_data_file=$TEST_FILE
+```
+
+This takes about half an hour to train on a single K80 GPU and about one minute for the evaluation to run. It reaches
+a score of ~20 perplexity once fine-tuned on the dataset.
+
+### RoBERTa/BERT and masked language modeling
+
+The following example fine-tunes RoBERTa on WikiText-2. Here too, we're using the raw WikiText-2. The loss is different
+as BERT/RoBERTa have a bidirectional mechanism; we're therefore using the same loss that was used during their
+pre-training: masked language modeling. 
+
+In accordance to the RoBERTa paper, we use dynamic masking rather than static masking. The model may, therefore, converge
+slightly slower (over-fitting takes more epochs).
+
+We use the `--mlm` flag so that the script may change its loss function.
+
+```bash
+export TRAIN_FILE=/path/to/dataset/wiki.train.raw
+export TEST_FILE=/path/to/dataset/wiki.test.raw
+
+python run_lm_finetuning.py \
+    --output_dir=output \
+    --model_type=roberta \
+    --model_name_or_path=roberta-base \
+    --do_train \
+    --train_data_file=$TRAIN_FILE \
+    --do_eval \
+    --eval_data_file=$TEST_FILE \
+    --mlm
+```
+
+## Language generation
+
+Based on the script [`run_generation.py`](https://github.com/huggingface/pytorch-transformers/blob/master/examples/run_generation.py).
+
+Conditional text generation using the auto-regressive models of the library: GPT, GPT-2, Transformer-XL and XLNet.
+A similar script is used for our official demo [Write With Transfomer](https://transformer.huggingface.co), where you
+can try out the different models available in the library.
+
+Example usage:
+
+```bash
+python run_generation.py \
+    --model_type=gpt2 \
+    --model_name_or_path=gpt2
+```
+
+## GLUE
+
+Based on the script [`run_glue.py`](https://github.com/huggingface/pytorch-transformers/blob/master/examples/run_glue.py).
+
+Fine-tuning the library models for sequence classification on the GLUE benchmark: [General Language Understanding 
+Evaluation](https://gluebenchmark.com/). This script can fine-tune the following models: BERT, XLM, XLNet and RoBERTa. 
+
+GLUE is made up of a total of 9 different tasks. We get the following results on the dev set of the benchmark with an
+uncased  BERT base model (the checkpoint `bert-base-uncased`). All experiments ran on 8  V100 GPUs with a total train
+batch size of 24. Some of these tasks have a small dataset and training can lead to high variance in the results
+between different runs. We report the median on 5 runs (with different seeds) for each of the metrics.
+
+| Task  | Metric                       | Result      |
+|-------|------------------------------|-------------|
+| CoLA  | Matthew's corr               | 55.75       |
+| SST-2 | Accuracy                     | 92.09       |
+| MRPC  | F1/Accuracy                  | 90.48/86.27 |
+| STS-B | Person/Spearman corr.        | 89.03/88.64 |
+| QQP   | Accuracy/F1                  | 90.92/87.72 |
+| MNLI  | Matched acc./Mismatched acc. | 83.74/84.06 |
+| QNLI  | Accuracy                     | 91.07       |
+| RTE   | Accuracy                     | 68.59       |
+| WNLI  | Accuracy                     | 43.66       |
+
+Some of these results are significantly different from the ones reported on the test set
+of GLUE benchmark on the website. For QQP and WNLI, please refer to [FAQ #12](https://gluebenchmark.com/faq) on the webite.
+
+Before running anyone of these GLUE tasks you should download the
+[GLUE data](https://gluebenchmark.com/tasks) by running
+[this script](https://gist.github.com/W4ngatang/60c2bdb54d156a41194446737ce03e2e)
+and unpack it to some directory `$GLUE_DIR`.
+
+```bash
+export GLUE_DIR=/path/to/glue
+export TASK_NAME=MRPC
+
+python run_glue.py \
+  --model_type bert \
+  --model_name_or_path bert-base-cased \
+  --task_name $TASK_NAME \
+  --do_train \
+  --do_eval \
+  --do_lower_case \
+  --data_dir $GLUE_DIR/$TASK_NAME \
+  --max_seq_length 128 \
+  --per_gpu_train_batch_size 32 \
+  --learning_rate 2e-5 \
+  --num_train_epochs 3.0 \
+  --output_dir /tmp/$TASK_NAME/
+```
+
+where task name can be one of CoLA, SST-2, MRPC, STS-B, QQP, MNLI, QNLI, RTE, WNLI.
+
+The dev set results will be present within the text file `eval_results.txt` in the specified output_dir. 
+In case of MNLI, since there are two separate dev sets (matched and mismatched), there will be a separate 
+output folder called `/tmp/MNLI-MM/` in addition to `/tmp/MNLI/`.
+
+The code has not been tested with half-precision training with apex on any GLUE task apart from MRPC, MNLI, 
+CoLA, SST-2. The following section provides details on how to run half-precision training with MRPC. With that being 
+said, there shouldn’t be any issues in running half-precision training with the remaining GLUE tasks as well, 
+since the data processor for each task inherits from the base class DataProcessor.
+
+### MRPC
+
+#### Fine-tuning example
+
+The following examples fine-tune BERT on the Microsoft Research Paraphrase Corpus (MRPC) corpus and runs in less 
+than 10 minutes on a single K-80 and in 27 seconds (!) on single tesla V100 16GB with apex installed.
+
+Before running anyone of these GLUE tasks you should download the
+[GLUE data](https://gluebenchmark.com/tasks) by running
+[this script](https://gist.github.com/W4ngatang/60c2bdb54d156a41194446737ce03e2e)
+and unpack it to some directory `$GLUE_DIR`.
+
+```bash
+export GLUE_DIR=/path/to/glue
+
+python run_glue.py \
+  --model_type bert \
+  --model_name_or_path bert-base-cased \
+  --task_name MRPC \
+  --do_train \
+  --do_eval \
+  --do_lower_case \
+  --data_dir $GLUE_DIR/MRPC/ \
+  --max_seq_length 128 \
+  --per_gpu_train_batch_size 32 \
+  --learning_rate 2e-5 \
+  --num_train_epochs 3.0 \
+  --output_dir /tmp/mrpc_output/
+```
+
+Our test ran on a few seeds with [the original implementation hyper-
+parameters](https://github.com/google-research/bert#sentence-and-sentence-pair-classification-tasks) gave evaluation 
+results between 84% and 88%.
+
+#### Using Apex and mixed-precision
+
+Using Apex and 16 bit precision, the fine-tuning on MRPC only takes 27 seconds. First install 
+[apex](https://github.com/NVIDIA/apex), then run the following example:
+
+```bash
+export GLUE_DIR=/path/to/glue
+
+python run_glue.py \
+  --model_type bert \
+  --model_name_or_path bert-base-cased \
+  --task_name MRPC \
+  --do_train \
+  --do_eval \
+  --do_lower_case \
+  --data_dir $GLUE_DIR/MRPC/ \
+  --max_seq_length 128 \
+  --per_gpu_train_batch_size 32 \
+  --learning_rate 2e-5 \
+  --num_train_epochs 3.0 \
+  --output_dir /tmp/mrpc_output/ \
+  --fp16
+```
+
+#### Distributed training
+
+Here is an example using distributed training on 8 V100 GPUs. The model used is the BERT whole-word-masking and it
+reaches F1 > 92 on MRPC.
+
+```bash
+export GLUE_DIR=/path/to/glue
+
+python -m torch.distributed.launch \
+    --nproc_per_node 8 run_glue.py \
+    --model_type bert \
+    --model_name_or_path bert-base-cased \
+    --task_name MRPC \
+    --do_train \
+    --do_eval \
+    --do_lower_case \
+    --data_dir $GLUE_DIR/MRPC/ \
+    --max_seq_length 128 \
+    --per_gpu_train_batch_size 8 \
+    --learning_rate 2e-5 \
+    --num_train_epochs 3.0 \
+    --output_dir /tmp/mrpc_output/
+```
+
+Training with these hyper-parameters gave us the following results:
+
+```bash
+acc = 0.8823529411764706
+acc_and_f1 = 0.901702786377709
+eval_loss = 0.3418912578906332
+f1 = 0.9210526315789473
+global_step = 174
+loss = 0.07231863956341798
+```
+
+### MNLI
+
+The following example uses the BERT-large, uncased, whole-word-masking model and fine-tunes it on the MNLI task.
+
+```bash
+export GLUE_DIR=/path/to/glue
+
+python -m torch.distributed.launch \
+    --nproc_per_node 8 run_glue.py \
+    --model_type bert \
+    --model_name_or_path bert-base-cased \
+    --task_name mnli \
+    --do_train \
+    --do_eval \
+    --do_lower_case \
+    --data_dir $GLUE_DIR/MNLI/ \
+    --max_seq_length 128 \
+    --per_gpu_train_batch_size 8 \
+    --learning_rate 2e-5 \
+    --num_train_epochs 3.0 \
+    --output_dir output_dir \
+```
+
+The results  are the following:
+
+```bash
+***** Eval results *****
+  acc = 0.8679706601466992
+  eval_loss = 0.4911287787382479
+  global_step = 18408
+  loss = 0.04755385363816904
+
+***** Eval results *****
+  acc = 0.8747965825874695
+  eval_loss = 0.45516540421714036
+  global_step = 18408
+  loss = 0.04755385363816904
+```
+
+##Multiple Choice
+
+Based on the script [`run_multiple_choice.py`]().
+
+#### Fine-tuning on SWAG
+Download [swag](https://github.com/rowanz/swagaf/tree/master/data) data
+
+```
+#training on 4 tesla V100(16GB) GPUS
+export SWAG_DIR=/path/to/swag_data_dir
+python ./examples/single_model_scripts/run_multiple_choice.py \
+--model_type roberta \
+--task_name swag \
+--model_name_or_path roberta-base \
+--do_train \
+--do_eval \
+--do_lower_case \
+--data_dir $SWAG_DIR \
+--learning_rate 5e-5 \
+--num_train_epochs 3 \
+--max_seq_length 80 \
+--output_dir models_bert/swag_base \
+--per_gpu_eval_batch_size=16 \
+--per_gpu_train_batch_size=16 \
+--gradient_accumulation_steps 2 \
+--overwrite_output
+```
+Training with the defined hyper-parameters yields the following results:
+```
+***** Eval results *****
+eval_acc = 0.8338998300509847
+eval_loss = 0.44457291918821606
+```
+
+## SQuAD
+
+Based on the script [`run_squad.py`](https://github.com/huggingface/pytorch-transformers/blob/master/examples/run_squad.py).
+
+#### Fine-tuning on SQuAD
+
+This example code fine-tunes BERT on the SQuAD dataset. It runs in 24 min (with BERT-base) or 68 min (with BERT-large) 
+on a single tesla V100 16GB. The data for SQuAD can be downloaded with the following links and should be saved in a 
+$SQUAD_DIR directory.
+
+* [train-v1.1.json](https://rajpurkar.github.io/SQuAD-explorer/dataset/train-v1.1.json)
+* [dev-v1.1.json](https://rajpurkar.github.io/SQuAD-explorer/dataset/dev-v1.1.json)
+* [evaluate-v1.1.py](https://github.com/allenai/bi-att-flow/blob/master/squad/evaluate-v1.1.py)
+
+```bash
+export SQUAD_DIR=/path/to/SQUAD
+
+python run_squad.py \
+  --model_type bert \
+  --model_name_or_path bert-base-cased \
+  --do_train \
+  --do_eval \
+  --do_lower_case \
+  --train_file $SQUAD_DIR/train-v1.1.json \
+  --predict_file $SQUAD_DIR/dev-v1.1.json \
+  --per_gpu_train_batch_size 12 \
+  --learning_rate 3e-5 \
+  --num_train_epochs 2.0 \
+  --max_seq_length 384 \
+  --doc_stride 128 \
+  --output_dir /tmp/debug_squad/
+```
+
+Training with the previously defined hyper-parameters yields the following results:
+
+```bash
+f1 = 88.52
+exact_match = 81.22
+```
+
+#### Distributed training
+
+
+Here is an example using distributed training on 8 V100 GPUs and Bert Whole Word Masking uncased model to reach a F1 > 93 on SQuAD:
+
+```bash
+python -m torch.distributed.launch --nproc_per_node=8 run_squad.py \
+    --model_type bert \
+    --model_name_or_path bert-base-cased \
+    --do_train \
+    --do_eval \
+    --do_lower_case \
+    --train_file $SQUAD_DIR/train-v1.1.json \
+    --predict_file $SQUAD_DIR/dev-v1.1.json \
+    --learning_rate 3e-5 \
+    --num_train_epochs 2 \
+    --max_seq_length 384 \
+    --doc_stride 128 \
+    --output_dir ../models/wwm_uncased_finetuned_squad/ \
+    --per_gpu_train_batch_size 24 \
+    --gradient_accumulation_steps 12
+```
+
+Training with the previously defined hyper-parameters yields the following results:
+
+```bash
+f1 = 93.15
+exact_match = 86.91
+```
+
+This fine-tuneds model is available as a checkpoint under the reference
+`bert-large-uncased-whole-word-masking-finetuned-squad`.
+

examples/contrib/README.md

+# Community contributed examples
+
+This folder contains examples which are not actively maintained (mostly contributed by the community).
+
+Using these examples together with a recent version of the library usually requires to make small (sometimes big) adaptations to get the scripts working.

examples/single_model_scripts/run_openai_gpt.py → examples/contrib/run_openai_gpt.py

@@ -153,9 +153,11 @@ def main():
     # This loading functions also add new tokens and embeddings called `special tokens`
     # These new embeddings will be fine-tuned on the RocStories dataset
     special_tokens = ['_start_', '_delimiter_', '_classify_']
-    tokenizer = OpenAIGPTTokenizer.from_pretrained(args.model_name, special_tokens=special_tokens)
-    special_tokens_ids = list(tokenizer.convert_tokens_to_ids(token) for token in special_tokens)
-    model = OpenAIGPTDoubleHeadsModel.from_pretrained(args.model_name, num_special_tokens=len(special_tokens))
+    tokenizer = OpenAIGPTTokenizer.from_pretrained(args.model_name)
+    tokenizer.add_tokens(special_tokens)
+    special_tokens_ids = tokenizer.convert_tokens_to_ids(special_tokens)
+    model = OpenAIGPTDoubleHeadsModel.from_pretrained(args.model_name)
+    model.resize_token_embeddings(len(tokenizer))
     model.to(device)
 
     # Load and encode the datasets
@@ -221,7 +223,7 @@ def main():
             for step, batch in enumerate(tqdm_bar):
                 batch = tuple(t.to(device) for t in batch)
                 input_ids, mc_token_ids, lm_labels, mc_labels = batch
-                losses = model(input_ids, mc_token_ids, lm_labels, mc_labels)
+                losses = model(input_ids, mc_token_ids=mc_token_ids, lm_labels=lm_labels, mc_labels=mc_labels)
                 loss = args.lm_coef * losses[0] + losses[1]
                 loss.backward()
                 scheduler.step()
@@ -258,7 +260,7 @@ def main():
             batch = tuple(t.to(device) for t in batch)
             input_ids, mc_token_ids, lm_labels, mc_labels = batch
             with torch.no_grad():
-               _, mc_loss, _, mc_logits = model(input_ids, mc_token_ids, lm_labels, mc_labels)
+               _, mc_loss, _, mc_logits = model(input_ids, mc_token_ids=mc_token_ids, lm_labels=lm_labels, mc_labels=mc_labels)
 
             mc_logits = mc_logits.detach().cpu().numpy()
             mc_labels = mc_labels.to('cpu').numpy()

examples/single_model_scripts/run_transfo_xl.py → examples/contrib/run_transfo_xl.py

@@ -113,7 +113,7 @@ def main():
         with torch.no_grad():
             mems = None
             for idx, (data, target, seq_len) in enumerate(eval_iter):
-                ret = model(data, target, mems)
+                ret = model(data, lm_labels=target, mems=mems)
                 loss, _, mems = ret
                 loss = loss.mean()
                 total_loss += seq_len * loss.item()

examples/distillation/README.md

@@ -9,6 +9,12 @@ DistilBERT stands for Distillated-BERT. DistilBERT is a small, fast, cheap and l
 For more information on DistilBERT, please refer to our [detailed blog post](https://medium.com/huggingface/smaller-faster-cheaper-lighter-introducing-distilbert-a-distilled-version-of-bert-8cf3380435b5
 ).
 
+## Setup
+
+This part of the library has only be tested with Python3.6+. There are few specific dependencies to install before launching a distillation, you can install them with the command `pip install -r requirements.txt`. 
+
+**Important note:** The training scripts have been updated to support PyTorch v1.2.0 (there are breakings changes compared to v1.1.0). It is important to note that there is a small internal bug in the current version of PyTorch available on pip that causes a memory leak in our training/distillation. It has been recently fixed and will likely be integrated into the next release. For the moment, we recommend to [compile PyTorch from source](https://github.com/pytorch/pytorch#from-source). Please refer to [issue 1179](https://github.com/huggingface/pytorch-transformers/issues/1179) for more details.
+
 ## How to use DistilBERT
 
 PyTorch-Transformers includes two pre-trained DistilBERT models, currently only provided for English (we are investigating the possibility to train and release a multilingual version of DistilBERT):

examples/distillation/dataset.py

@@ -77,7 +77,7 @@ class Dataset:
                     if sub_s[0] != cls_id:
                         sub_s = np.insert(sub_s, 0, cls_id)
                     if sub_s[-1] != sep_id:
-                        sub_s = np.insert(sub_s, len(sub_s), cls_id)
+                        sub_s = np.insert(sub_s, len(sub_s), sep_id)
                     assert len(sub_s) <= max_len
                     sub_seqs.append(sub_s)
 

examples/distillation/distiller.py

@@ -17,6 +17,7 @@
 """
 import os
 import math
+import psutil
 from tensorboardX import SummaryWriter
 from tqdm import trange, tqdm
 import numpy as np
@@ -192,7 +193,7 @@ class Distiller:
         x_prob = self.token_probs[token_ids.flatten()]
         n_tgt = math.ceil(self.mlm_mask_prop * lengths.sum().item())
         tgt_ids = torch.multinomial(x_prob / x_prob.sum(), n_tgt, replacement=False)
-        pred_mask = torch.zeros(bs * max_seq_len, dtype=torch.uint8, device=token_ids.device)
+        pred_mask = torch.zeros(bs * max_seq_len, dtype=torch.bool, device=token_ids.device) # previously `dtype=torch.uint8`, cf pytorch 1.2.0 compatibility
         pred_mask[tgt_ids] = 1
         pred_mask = pred_mask.view(bs, max_seq_len)
 
@@ -216,7 +217,7 @@ class Distiller:
         _token_ids = _token_ids_mask * (probs == 0).long() + _token_ids_real * (probs == 1).long() + _token_ids_rand * (probs == 2).long()
         token_ids = token_ids.masked_scatter(pred_mask, _token_ids)
 
-        mlm_labels[1-pred_mask] = -1
+        mlm_labels[~pred_mask] = -1 # previously `mlm_labels[1-pred_mask] = -1`, cf pytorch 1.2.0 compatibility
 
         return token_ids, attn_mask, mlm_labels
 
@@ -294,7 +295,10 @@ class Distiller:
             if self.is_master: logger.info(f'--- Ending epoch {self.epoch}/{self.params.n_epoch-1}')
             self.end_epoch()
 
-        if self.is_master: logger.info('Training is finished')
+        if self.is_master:
+            logger.info(f'Save very last checkpoint as `pytorch_model.bin`.')
+            self.save_checkpoint(checkpoint_name=f'pytorch_model.bin')
+            logger.info('Training is finished')
 
     def step(self,
              input_ids: torch.tensor,
@@ -379,9 +383,9 @@ class Distiller:
                 torch.nn.utils.clip_grad_norm_(amp.master_params(self.optimizer), self.params.max_grad_norm)
             else:
                 torch.nn.utils.clip_grad_norm_(self.student.parameters(), self.params.max_grad_norm)
-            self.scheduler.step()
             self.optimizer.step()
             self.optimizer.zero_grad()
+            self.scheduler.step()
 
     def iter(self):
         """
@@ -419,6 +423,8 @@ class Distiller:
             self.tensorboard.add_scalar(tag="losses/loss_mse", scalar_value=self.last_loss_mse, global_step=self.n_total_iter)
         self.tensorboard.add_scalar(tag="learning_rate/lr", scalar_value=self.scheduler.get_lr()[0], global_step=self.n_total_iter)
         
+        self.tensorboard.add_scalar(tag="global/memory_usage", scalar_value=psutil.virtual_memory()._asdict()['used']/1_000_000, global_step=self.n_total_iter)
+
     def end_epoch(self):
         """
         Finally arrived at the end of epoch (full pass on dataset).

examples/distillation/requirements.txt

 gitpython==3.0.2
+tensorboard>=1.14.0
+tensorboardX==1.8
+psutil==5.6.3

examples/distillation/scripts/binarized_data.py

@@ -21,8 +21,12 @@ import random
 import time
 import numpy as np
 from pytorch_transformers import BertTokenizer
+import logging
 
-from examples.distillation.utils import logger
+logging.basicConfig(format = '%(asctime)s - %(levelname)s - %(name)s -   %(message)s',
+                    datefmt = '%m/%d/%Y %H:%M:%S',
+                    level = logging.INFO)
+logger = logging.getLogger(__name__)
 
 def main():
     parser = argparse.ArgumentParser(description="Preprocess the data to avoid re-doing it several times by (tokenization + token_to_ids).")

examples/distillation/scripts/token_counts.py

@@ -18,8 +18,12 @@ Preprocessing script before training DistilBERT.
 from collections import Counter
 import argparse
 import pickle
+import logging
 
-from examples.distillation.utils import logger
+logging.basicConfig(format = '%(asctime)s - %(levelname)s - %(name)s -   %(message)s',
+                    datefmt = '%m/%d/%Y %H:%M:%S',
+                    level = logging.INFO)
+logger = logging.getLogger(__name__)
 
 if __name__ == '__main__':
     parser = argparse.ArgumentParser(description="Token Counts for smoothing the masking probabilities in MLM (cf XLM/word2vec)")

examples/lm_finetuning/README.md

-# BERT Model Finetuning using Masked Language Modeling objective
-
-## Introduction
-
-The three example scripts in this folder can be used to **fine-tune** a pre-trained BERT model using the pretraining objective (combination of masked language modeling and next sentence prediction loss). In general, pretrained models like BERT are first trained with a pretraining objective (masked language modeling and next sentence prediction for BERT) on a large and general natural language corpus. A classifier head is then added on top of the pre-trained architecture and the model is quickly fine-tuned on a target task, while still (hopefully) retaining its general language understanding. This greatly reduces overfitting and yields state-of-the-art results, especially when training data for the target task are limited.
-
-The [ULMFiT paper](https://arxiv.org/abs/1801.06146) took a slightly different approach, however, and added an intermediate step in which the model is fine-tuned on text **from the same domain as the target task and using the pretraining objective** before the final stage in which the classifier head is added and the model is trained on the target task itself. This paper reported significantly improved results from this step, and found that they could get high-quality classifications even with only tiny numbers (<1000) of labelled training examples, as long as they had a lot of unlabelled data from the target domain.
-
-Although this wasn't covered in the original BERT paper, domain-specific fine-tuning of Transformer models has [recently been reported by other authors](https://arxiv.org/pdf/1905.05583.pdf), and they report performance improvements as well.
-
-## Input format
-
-The scripts in this folder expect a single file as input, consisting of untokenized text, with one **sentence** per line, and one blank line between documents. The reason for the sentence splitting is that part of BERT's training involves a _next sentence_ objective in which the model must predict whether two sequences of text are contiguous text from the same document or not, and to avoid making the task _too easy_, the split point between the sequences is always at the end of a sentence. The linebreaks in the file are therefore necessary to mark the points where the text can be split.
-
-## Usage
-
-There are two ways to fine-tune a language model using these scripts. The first _quick_ approach is to use [`simple_lm_finetuning.py`](./simple_lm_finetuning.py). This script does everything in a single script, but generates training instances that consist of just two sentences. This is quite different from the BERT paper, where (confusingly) the NextSentence task concatenated sentences together from each document to form two long multi-sentences, which the paper just referred to as _sentences_. The difference between this simple approach and the original paper approach can have a significant effect for long sequences since two sentences will be much shorter than the max sequence length. In this case, most of each training example will just consist of blank padding characters, which wastes a lot of computation and results in a model that isn't really training on long sequences.
-
-As such, the preferred approach (assuming you have documents containing multiple contiguous sentences from your target domain) is to use [`pregenerate_training_data.py`](./pregenerate_training_data.py) to pre-process your data into training examples following the methodology used for LM training in the original BERT paper and repository. Since there is a significant random component to training data generation for BERT, this script includes an option to generate multiple _epochs_ of pre-processed data, to avoid training on the same random splits each epoch. Generating an epoch of data for each training epoch should result a better final model, and so we recommend doing so.
-
-You can then train on the pregenerated data using [`finetune_on_pregenerated.py`](./finetune_on_pregenerated.py), and pointing it to the folder created by [`pregenerate_training_data.py`](./pregenerate_training_data.py). Note that you should use the same `bert_model` and case options for both! Also note that `max_seq_len` does not need to be specified for the [`finetune_on_pregenerated.py`](./finetune_on_pregenerated.py) script, as it is inferred from the training examples.
-
-There are various options that can be tweaked, but they are mostly set to the values from the BERT paper/repository and default values should make sense. The most relevant ones are:
-
-- `--max_seq_len`: Controls the length of training examples (in wordpiece tokens) seen by the model. Defaults to 128 but can be set as high as 512. Higher values may yield stronger language models at the cost of slower and more memory-intensive training.
-- `--fp16`: Enables fast half-precision training on recent GPUs.
-
-In addition, if memory usage is an issue, especially when training on a single GPU, reducing `--train_batch_size` from the default 32 to a lower number (4-16) can be helpful, or leaving `--train_batch_size` at the default and increasing `--gradient_accumulation_steps` to 2-8. Changing `--gradient_accumulation_steps` may be preferable as alterations to the batch size may require corresponding changes in the learning rate to compensate. There is also a `--reduce_memory` option for both the `pregenerate_training_data.py` and `finetune_on_pregenerated.py` scripts that spills data to disc in shelf objects or numpy memmaps rather than retaining it in memory, which significantly reduces memory usage with little performance impact.
-
-## Examples
-
-### Simple fine-tuning
-
-```
-python3 simple_lm_finetuning.py 
---train_corpus my_corpus.txt 
---bert_model bert-base-uncased 
---do_lower_case 
---output_dir finetuned_lm/
---do_train
-```
-
-### Pregenerating training data
-
-```
-python3 pregenerate_training_data.py
---train_corpus my_corpus.txt
---bert_model bert-base-uncased
---do_lower_case
---output_dir training/
---epochs_to_generate 3
---max_seq_len 256
-```
-
-### Training on pregenerated data
-
-```
-python3 finetune_on_pregenerated.py
---pregenerated_data training/
---bert_model bert-base-uncased
---do_lower_case
---output_dir finetuned_lm/
---epochs 3
-```