Create README.md

README.md

@@ -0,0 +1,73 @@
+---
+language: ja
+license: cc-by-sa-4.0
+library_name: transformers
+tags:
+  - gpt2
+datasets:
+  - wikipedia
+  - cc100
+  - oscar
+widget:
+    - text: "<s>昨日私は京都で"
+---
+
+# Model Card for Japanese character-level GPT-2 Large
+
+## Model description
+
+This is a Japanese character-level GPT-2 Large (717M parameters) language model pre-trained on Japanese Wikipedia, the Japanese portion of CC-100, and the Japanese portion of OSCAR.
+
+## How to use
+
+You can use this model directly with a pipeline for text generation.
+
+```python
+>>> from transformers import pipeline, set_seed
+>>> generator = pipeline('text-generation', model='ku-nlp/gpt2-large-japanese-char')
+>>> set_seed(5)
+>>> generator("<s>昨日私は京都で", max_length=30, do_sample=True, num_return_sequences=5)
+
+[{'generated_text': '<s>昨日私は京都で仕事だったのですが、帰りは車を信号で止めて、'},
+ {'generated_text': '<s>昨日私は京都で開かれた大阪市都市戦略会議に出席しました。そ'},
+ {'generated_text': '<s>昨日私は京都で行われました関西の教育者・学校事例が集まるイ'},
+ {'generated_text': '<s>昨日私は京都では初雪を見ました。朝は少しパッとしない天気で'},
+ {'generated_text': '<s>昨日私は京都でこみっくトレジャーさんの撮影を見学させていた'}]
+```
+
+You can also use this model to get the features of a given text.
+
+## Vocabulary
+
+A character-level vocabulary of size 6K is used. To be precise, rare characters may be split into bytes because byte-level byte-pair encoding (BPE) is used. The BPE tokenizer was trained on a small subset of the training data. Since the data were converted into a one-character-per-line format, merge operations never go beyond character boundaries.
+
+Note that the tokenizer maps U+0020 to `[UNK]` because preprocessing eliminated whitespace characters (U+0020) from training data. Use U+3000 (Ideographic Space) instead.
+
+## Training data
+
+We used the following corpora for pre-training:
+
+- Japanese Wikipedia (as of 20221020, 3.2GB, 27M sentences, 1.3M documents)
+- Japanese portion of CC-100 (85GB, 619M sentences, 66M documents)
+- Japanese portion of OSCAR (54GB, 326M sentences, 25M documents)
+
+Note that we filtered out documents annotated with "header", "footer", or "noisy" tags in OSCAR.
+Also note that Japanese Wikipedia was duplicated 10 times to make the total size of the corpus comparable to that of CC-100 and OSCAR. As a result, the total size of the training data is 171GB.
+
+## Training procedure
+
+The training took about 8 months (with 7 interruptions) with a single NVIDIA A100 80GB GPU.
+
+The following hyperparameters were used during pre-training:
+
+- learning_rate: 2e-4
+- per_device_train_batch_size: 6
+- gradient_accumulation_steps: 98
+- optimizer: AdamW with betas=(0.9, 0.999) and epsilon=1e-06
+- weight_decay: 0.01
+- lr_scheduler_type: linear
+- max_grad_norm: 1.0
+- max_steps: 500,000 (but terminated at 186,000 steps ~= 2.0 epochs)
+- warmup_steps: 10,000
+
+The eval loss was 1.309 while the eval accuracy was 0.6890. The evaluation set consists of 5,000 randomly sampled documents from each of the training corpora.