Upload 9 files

arm_tokenizer/LICENSE.txt

@@ -0,0 +1,201 @@
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following
+      boilerplate notice, with the fields enclosed by brackets "[]"
+      replaced with your own identifying information. (Don't include
+      the brackets!)  The text should be enclosed in the appropriate
+      comment syntax for the file format. We also recommend that a
+      file or class name and description of purpose be included on the
+      same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+   Copyright 2026 NeoneAI
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.

arm_tokenizer/README.md

@@ -0,0 +1,87 @@
+---
+license: apache-2.0
+language:
+  - ru
+  - en
+tags:
+  - tokenizer
+  - bpe
+  - byte-level
+  - russian
+  - english
+  - emoji
+---
+
+# ARM Tokenizer
+
+A byte-level BPE tokenizer built by **NeoneAI** for Russian, English, and emoji.
+
+Part of the ARM research project.
+
+## Overview
+
+- **Vocab size:** 65 536
+- **Algorithm:** Byte-level BPE
+- **Languages:** Russian + English
+- **Emoji:** supported
+- **Special tokens:** `<|endoftext|>`, `<|pad|>`, `<|user|>`, `<|assistant|>`, `<|system|>`, `<|bos|>`, `<|eos|>`
+- **Tool:** Hugging Face Tokenizers
+
+## Why
+
+Most open tokenizers (GPT-2, LLaMA, Mistral) are English-first.
+ARM Tokenizer is built for **Russian + English + emoji**.
+
+It compresses Russian text **3.3× better** than GPT-2
+and beats Gemma-2 (256k vocab) with only a 65k vocabulary.
+
+## Benchmarks
+
+All benchmarks use **1000 randomly sampled real-world texts**
+(500 Russian + 500 English) with fixed seed (`42`).
+Margin of error: **±1%**.
+
+### vs GPT-2 (50k)
+
+| Metric | ARM | GPT-2 |
+|---|---|---|
+| Total tokens | **153 473** | 414 971 |
+| Avg tok/char | **0.2965** | 0.8018 |
+| Russian (500) | **111 803** | 371 960 |
+| English (500) | **41 670** | 43 011 |
+| Wins | **928** | 16 |
+| Ties | 56 | |
+
+**Result:** ARM is **63.02% ±1% more efficient** than GPT-2.
+On Russian, ARM is **3.3× more efficient** than GPT-2.
+
+### vs Gemma-2 (256k)
+
+| Metric | ARM | Gemma-2 |
+|---|---|---|
+| Total tokens | **153 473** | 167 105 |
+| Avg tok/char | **0.2965** | 0.3198 |
+| Russian (500) | **111 803** | 126 941 |
+| English (500) | 41 670 | **40 164** |
+| Wins | **791** | 108 |
+| Ties | 101 | |
+
+**Result:** ARM is **8.16% ±1% more efficient** than Gemma-2 overall,
+and **12% ±1% more efficient on Russian**, with only a 65k vocabulary.
+
+## Reproducing the benchmarks
+
+Two test scripts are included:
+
+- **`test_gpt-2.py`** — compares ARM Tokenizer with GPT-2 Tokenizer
+- **`test_gemma-2.py`** — compares ARM Tokenizer with Gemma-2 Tokenizer
+
+Both scripts read from `sample/` folder:
+
+- `sample/gutenberg_sample.txt` — 5000 English texts from Project Gutenberg
+- `sample/ru_books_sample.txt` — 5000 Russian texts from RuHeritage-Corpus
+
+### Generate test data
+
+```bash
+python build_test_data.py

arm_tokenizer/README.ru.md

@@ -0,0 +1,93 @@
+
+---
+
+## README.ru.md (Русский)
+
+```markdown
+---
+license: apache-2.0
+language:
+  - ru
+  - en
+tags:
+  - tokenizer
+  - bpe
+  - byte-level
+  - russian
+  - english
+  - emoji
+---
+
+# ARM Tokenizer
+
+Байтовый BPE-токенизатор от **NeoneAI** для русского, английского и эмодзи.
+
+Часть исследовательского проекта ARM.
+
+## Обзор
+
+- **Размер словаря:** 65 536
+- **Алгоритм:** Byte-level BPE
+- **Языки:** русский + английский
+- **Эмодзи:** поддерживаются
+- **Спецтокены:** `<|endoftext|>`, `<|pad|>`, `<|user|>`, `<|assistant|>`, `<|system|>`, `<|bos|>`, `<|eos|>`
+- **Инструмент:** Hugging Face Tokenizers
+
+## Зачем
+
+Большинство открытых токенизаторов (GPT-2, LLaMA, Mistral) заточены под английский.
+ARM Tokenizer создан для **русского + английского + эмодзи**.
+
+Он сжимает русский текст **в 3.3 раза лучше**, чем GPT-2,
+и обходит Gemma-2 (256k словарь) при словаре всего 65k.
+
+## Результаты
+
+Все тесты на **1000 случайных реальных текстов**
+(500 русских + 500 английских) с фиксированным seed (`42`).
+Погрешность: **±1%**.
+
+### vs GPT-2 (50k)
+
+| Метрика | ARM | GPT-2 |
+|---|---|---|
+| Всего токенов | **153 473** | 414 971 |
+| Средний tok/char | **0.2965** | 0.8018 |
+| Русский (500) | **111 803** | 371 960 |
+| Английский (500) | **41 670** | 43 011 |
+| Побед | **928** | 16 |
+| Ничьих | 56 | |
+
+**Итог:** ARM на **63.02% ±1% эффективнее** GPT-2.
+На русском ARM в **3.3 раза эффективнее** GPT-2.
+
+### vs Gemma-2 (256k)
+
+| Метрика | ARM | Gemma-2 |
+|---|---|---|
+| Всего токенов | **153 473** | 167 105 |
+| Средний tok/char | **0.2965** | 0.3198 |
+| Русский (500) | **111 803** | 126 941 |
+| Английский (500) | 41 670 | **40 164** |
+| Побед | **791** | 108 |
+| Ничьих | 101 | |
+
+**Итог:** ARM на **8.16% ±1% эффективнее** Gemma-2 в общем,
+и **на 12% ±1% эффективнее на русском**, при словаре всего 65k.
+
+## Воспроизведение тестов
+
+В репозитории два скрипта:
+
+- **`test_gpt-2.py`** — сравнивает ARM Tokenizer с GPT-2 Tokenizer
+- **`test_gemma-2.py`** — сравнивает ARM Tokenizer с Gemma-2 Tokenizer
+
+Оба скрипта читают из папки `sample/`:
+
+- `sample/gutenberg_sample.txt` — 5000 английских текстов из Project Gutenberg
+- `sample/ru_books_sample.txt` — 5000 русских текстов из RuHeritage-Corpus
+
+### Создать тестовые данные
+
+```bash
+python build_test_data.py

arm_tokenizer/build_test_data.py

@@ -0,0 +1,95 @@
+import os
+import random
+from pathlib import Path
+from datasets import load_dataset
+
+GUTENBERG = "D:/gutenberg_en.txt"
+RU_BOOKS  = "D:/ru_books.txt"
+
+SCRIPT_DIR = Path(__file__).parent
+OUT_DIR = SCRIPT_DIR / "sample"
+OUT_DIR.mkdir(parents=True, exist_ok=True)
+
+OUT_EN = OUT_DIR / "gutenberg_sample.txt"
+OUT_RU = OUT_DIR / "ru_books_sample.txt"
+
+N_SAMPLES = 5000
+MIN_LEN   = 100
+MAX_LEN   = 2000
+MAX_SCAN  = 2_000_000
+
+
+def download_gutenberg():
+    print("downloading gutenberg...")
+    ds = load_dataset("AdhyanshVerma/pg-en", split="train", streaming=True)
+    with open(GUTENBERG, "w", encoding="utf-8") as f:
+        n = 0
+        for s in ds:
+            t = s.get("text", "").strip()
+            if t:
+                f.write(t + "\n")
+                n += 1
+            if n >= 500_000:
+                break
+    print(f"gutenberg: {n} lines")
+
+
+def download_ru_books():
+    print("downloading ru books...")
+    ds = load_dataset("maxzt/RuHeritage-Corpus", split="train", streaming=True)
+    with open(RU_BOOKS, "w", encoding="utf-8") as f:
+        n = 0
+        for s in ds:
+            t = s.get("text", "").strip()
+            if t:
+                f.write(t + "\n")
+                n += 1
+            if n >= 500_000:
+                break
+    print(f"ru books: {n} lines")
+
+
+def make_sample(src_path, dst_path, n_samples, min_len, max_len):
+    if not os.path.exists(src_path):
+        print(f"missing: {src_path}")
+        return 0
+
+    candidates = []
+    total = 0
+
+    with open(src_path, "r", encoding="utf-8", errors="ignore") as f:
+        for line in f:
+            total += 1
+            line = line.strip()
+            if min_len <= len(line) <= max_len:
+                candidates.append(line)
+            if total >= MAX_SCAN:
+                break
+
+    if not candidates:
+        return 0
+
+    random.shuffle(candidates)
+    selected = candidates[:n_samples]
+
+    with open(dst_path, "w", encoding="utf-8") as f:
+        for line in selected:
+            f.write(line + "\n")
+
+    return len(selected)
+
+
+if __name__ == "__main__":
+    random.seed(42)
+
+    if not os.path.exists(GUTENBERG):
+        download_gutenberg()
+
+    if not os.path.exists(RU_BOOKS):
+        download_ru_books()
+
+    en = make_sample(GUTENBERG, OUT_EN, N_SAMPLES, MIN_LEN, MAX_LEN)
+    ru = make_sample(RU_BOOKS, OUT_RU, N_SAMPLES, MIN_LEN, MAX_LEN)
+
+    print(f"en: {en}")
+    print(f"ru: {ru}")

arm_tokenizer/test_gemma-2.py

@@ -0,0 +1,104 @@
+import os
+import random
+from tokenizers import Tokenizer
+from transformers import AutoTokenizer
+
+SCRIPT_DIR = os.path.dirname(os.path.abspath(__file__))
+SAMPLE_DIR = os.path.join(SCRIPT_DIR, "sample")
+
+GUTENBERG = os.path.join(SAMPLE_DIR, "gutenberg_sample.txt")
+RU_BOOKS  = os.path.join(SAMPLE_DIR, "ru_books_sample.txt")
+
+TOKENIZER_PATH = os.path.join(SCRIPT_DIR, "tokenizer.json")
+
+N = 500
+MIN_LEN = 100
+MAX_LEN = 2000
+
+arm = Tokenizer.from_file(TOKENIZER_PATH)
+gemma = AutoTokenizer.from_pretrained("google/gemma-2-2b", trust_remote_code=True)
+
+
+def read(path, n):
+    if not os.path.exists(path):
+        print(f"missing: {path}")
+        return []
+    s = []
+    with open(path, "r", encoding="utf-8") as f:
+        for line in f:
+            line = line.strip()
+            if MIN_LEN <= len(line) <= MAX_LEN:
+                s.append(line)
+    random.shuffle(s)
+    return s[:n]
+
+
+random.seed(42)
+
+en = read(GUTENBERG, N)
+ru = read(RU_BOOKS, N)
+
+samples = [(t, "en") for t in en] + [(t, "ru") for t in ru]
+
+print(f"en: {len(en)}")
+print(f"ru: {len(ru)}")
+print(f"total: {len(samples)}")
+
+total_arm = 0
+total_gemma = 0
+arm_wins = 0
+gemma_wins = 0
+ties = 0
+
+by_lang = {
+    "en": {"arm": 0, "gemma": 0, "chars": 0, "n": 0},
+    "ru": {"arm": 0, "gemma": 0, "chars": 0, "n": 0},
+}
+
+for text, lang in samples:
+    a = len(arm.encode(text).ids)
+    g = len(gemma.encode(text))
+
+    total_arm += a
+    total_gemma += g
+
+    by_lang[lang]["arm"] += a
+    by_lang[lang]["gemma"] += g
+    by_lang[lang]["chars"] += len(text)
+    by_lang[lang]["n"] += 1
+
+    if a < g:
+        arm_wins += 1
+    elif a > g:
+        gemma_wins += 1
+    else:
+        ties += 1
+
+print("=" * 60)
+print(f"ARM:   {total_arm:,}")
+print(f"Gemma: {total_gemma:,}")
+
+print("=" * 60)
+for lang, d in by_lang.items():
+    if d["n"] == 0:
+        continue
+    ar = d["arm"] / d["chars"]
+    gr = d["gemma"] / d["chars"]
+    diff = d["arm"] - d["gemma"]
+    print(f"{lang}: ARM={d['arm']:,} Gemma={d['gemma']:,} tok/char ARM={ar:.4f} Gemma={gr:.4f} delta={diff:+,}")
+
+print("=" * 60)
+print(f"ARM wins:    {arm_wins}")
+print(f"Gemma wins:  {gemma_wins}")
+print(f"ties:        {ties}")
+
+diff = total_arm - total_gemma
+pct = abs(diff) / max(total_arm, total_gemma) * 100
+
+print("=" * 60)
+if diff < 0:
+    print(f"WINNER: ARM by {pct:.2f}%")
+elif diff > 0:
+    print(f"WINNER: Gemma by {pct:.2f}%")
+else:
+    print("WINNER: TIE")

arm_tokenizer/test_gpt-2.py

@@ -0,0 +1,98 @@
+import os
+import random
+from tokenizers import Tokenizer
+from transformers import GPT2TokenizerFast
+
+GUTENBERG = "sample/gutenberg_sample.txt"
+RU_BOOKS  = "sample/ru_books_sample.txt"
+
+N = 500
+MIN_LEN = 100
+MAX_LEN = 2000
+
+arm = Tokenizer.from_file("tokenizer.json")
+gpt2 = GPT2TokenizerFast.from_pretrained("gpt2")
+
+
+def read(path, n):
+    if not os.path.exists(path):
+        return []
+    s = []
+    with open(path, "r", encoding="utf-8") as f:
+        for line in f:
+            line = line.strip()
+            if MIN_LEN <= len(line) <= MAX_LEN:
+                s.append(line)
+    random.shuffle(s)
+    return s[:n]
+
+
+random.seed(42)
+
+en = read(GUTENBERG, N)
+ru = read(RU_BOOKS, N)
+
+samples = [(t, "en") for t in en] + [(t, "ru") for t in ru]
+
+print(f"en: {len(en)}")
+print(f"ru: {len(ru)}")
+print(f"total: {len(samples)}")
+
+total_arm = 0
+total_gpt2 = 0
+arm_wins = 0
+gpt2_wins = 0
+ties = 0
+
+by_lang = {
+    "en": {"arm": 0, "gpt2": 0, "chars": 0, "n": 0},
+    "ru": {"arm": 0, "gpt2": 0, "chars": 0, "n": 0},
+}
+
+for text, lang in samples:
+    a = len(arm.encode(text).ids)
+    g = len(gpt2.encode(text))
+
+    total_arm += a
+    total_gpt2 += g
+
+    by_lang[lang]["arm"] += a
+    by_lang[lang]["gpt2"] += g
+    by_lang[lang]["chars"] += len(text)
+    by_lang[lang]["n"] += 1
+
+    if a < g:
+        arm_wins += 1
+    elif a > g:
+        gpt2_wins += 1
+    else:
+        ties += 1
+
+print("=" * 60)
+print(f"ARM-300: {total_arm:,}")
+print(f"GPT-2:   {total_gpt2:,}")
+
+print("=" * 60)
+for lang, d in by_lang.items():
+    if d["n"] == 0:
+        continue
+    ar = d["arm"] / d["chars"]
+    gr = d["gpt2"] / d["chars"]
+    diff = d["arm"] - d["gpt2"]
+    print(f"{lang}: ARM={d['arm']:,} GPT2={d['gpt2']:,} tok/char ARM={ar:.4f} GPT2={gr:.4f} delta={diff:+,}")
+
+print("=" * 60)
+print(f"ARM wins:   {arm_wins}")
+print(f"GPT2 wins:  {gpt2_wins}")
+print(f"ties:       {ties}")
+
+diff = total_arm - total_gpt2
+pct = abs(diff) / max(total_arm, total_gpt2) * 100
+
+print("=" * 60)
+if diff < 0:
+    print(f"WINNER: ARM-300 by {pct:.2f}%")
+elif diff > 0:
+    print(f"WINNER: GPT-2 by {pct:.2f}%")
+else:
+    print("WINNER: TIE")