Update README.md

Browse files

Files changed (1) hide show

README.md +106 -184

README.md CHANGED Viewed

@@ -1,199 +1,121 @@
 ---
 library_name: transformers
-tags: []
 ---
-# Model Card for Model ID
-<!-- Provide a quick summary of what the model is/does. -->
-## Model Details
-### Model Description
-<!-- Provide a longer summary of what this model is. -->
-This is the model card of a 🤗 transformers model that has been pushed on the Hub. This model card has been automatically generated.
-- **Developed by:** [More Information Needed]
-- **Funded by [optional]:** [More Information Needed]
-- **Shared by [optional]:** [More Information Needed]
-- **Model type:** [More Information Needed]
-- **Language(s) (NLP):** [More Information Needed]
-- **License:** [More Information Needed]
-- **Finetuned from model [optional]:** [More Information Needed]
-### Model Sources [optional]
-<!-- Provide the basic links for the model. -->
-- **Repository:** [More Information Needed]
-- **Paper [optional]:** [More Information Needed]
-- **Demo [optional]:** [More Information Needed]
-## Uses
-<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
-### Direct Use
-<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
-[More Information Needed]
-### Downstream Use [optional]
-<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
-[More Information Needed]
-### Out-of-Scope Use
-<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
-[More Information Needed]
-## Bias, Risks, and Limitations
-<!-- This section is meant to convey both technical and sociotechnical limitations. -->
-[More Information Needed]
-### Recommendations
-<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
-Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
-## How to Get Started with the Model
-Use the code below to get started with the model.
-[More Information Needed]
-## Training Details
-### Training Data
-<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
-[More Information Needed]
-### Training Procedure
-<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
-#### Preprocessing [optional]
-[More Information Needed]
-#### Training Hyperparameters
-- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
-#### Speeds, Sizes, Times [optional]
-<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
-[More Information Needed]
-## Evaluation
-<!-- This section describes the evaluation protocols and provides the results. -->
-### Testing Data, Factors & Metrics
-#### Testing Data
-<!-- This should link to a Dataset Card if possible. -->
-[More Information Needed]
-#### Factors
-<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
-[More Information Needed]
-#### Metrics
-<!-- These are the evaluation metrics being used, ideally with a description of why. -->
-[More Information Needed]
-### Results
-[More Information Needed]
-#### Summary
-## Model Examination [optional]
-<!-- Relevant interpretability work for the model goes here -->
-[More Information Needed]
-## Environmental Impact
-<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
-Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
-- **Hardware Type:** [More Information Needed]
-- **Hours used:** [More Information Needed]
-- **Cloud Provider:** [More Information Needed]
-- **Compute Region:** [More Information Needed]
-- **Carbon Emitted:** [More Information Needed]
-## Technical Specifications [optional]
-### Model Architecture and Objective
-[More Information Needed]
-### Compute Infrastructure
-[More Information Needed]
-#### Hardware
-[More Information Needed]
-#### Software
-[More Information Needed]
-## Citation [optional]
-<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
-**BibTeX:**
-[More Information Needed]
-**APA:**
-[More Information Needed]
-## Glossary [optional]
-<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
-[More Information Needed]
-## More Information [optional]
-[More Information Needed]
-## Model Card Authors [optional]
-[More Information Needed]
-## Model Card Contact
-[More Information Needed]

 ---
 library_name: transformers
+tags:
+- avito
+- multimodal
+- vlm
+- vision-language
+- ocr
+license: apache-2.0
+language:
+- ru
+- en
+base_model:
+- Qwen/Qwen2.5-VL-7B-Instruct
+pipeline_tag: image-text-to-text
 ---
+# A-Vision — русскоязычная VLM Авито
+A-Vision — Visual-Language модель, адаптированная под русский язык и домен Авито. Она понимает изображение и текст вместе: описывает фото, отвечает на вопросы по картинке, сверяет соответствие описания и фото, извлекает бренды/надписи/произвольный текст (OCR).
+## Зачем и как делали
+* **Данные.** Собрали собственный русскоязычный мультимодальный корпус: ~200k изображений объявлений и ≈1M пар «вопрос–ответ», дополненный тщательно локализованными наборами (вместо «сырого» машинного перевода).Также перевели несколько OS-датасетов.
+* **Адаптация LLM.** Заменили токенизатор на русскоязычный; провели **freeze→unfreeze** LLM-части модели на большом корпусе русскоязычного текста.
+* **Мультимодальное SFT.** Дообучили модель на собранном датасете «изображение+вопрос → ответ».
+* **RL-этап.** Проверили DPO, которое позволило добиться от модели безопасных ответов.
+* **Результат.** Рост качества на русскоязычных и доменных тестах (Авито-метрика генерации описаний +5.6%, MMMU_RU +2.6%, RealWorldQA_RU +1.9%) при сохранении универсальных VLM-навыков; небольшая просадка на части англоязычных бенчмарков ожидаема из-за фокуса на русском.
+| Метрика          | Qwen2.5-VL-7B-Instruct | **A-Vision** |
+| :--------------- | :--------------------: | :----------: |
+| AvitoImageGen_RU |         0.7259         |  **0.7668**  |
+| MMMU_EN          |        **0.543**       |     0.489    |
+| MMMU_RU          |          0.469         |   **0.474**  |
+| RealWorldQA_EN   |          0.673         |   **0.693**  |
+| RealWorldQA_RU   |          0.647         |   **0.652**  |
+| OCRBench_EN      |         **878**        |      834     |
+| OCRVQA_EN        |       **77.506**       |    74.4098   |
+| ChartQA_EN       |        **86.44**       |      86      |
+| DocVQA_EN        |         94.7458        |  **94.9702** |
+В токенизаторе A-vision плотность токенизации выше, чем у Qwen2.5-VL-7B-Instruct поэтому число токенов в контексте и при генерации стало меньше для одинаковых примеров. Кроме того, размер самой модели сократился до 7.9B при 8.2B у Qwen3-8B. За счет этого одинаковые русскоязычные примеры адаптированной моделью обрабатываются быстрее в среднем на 15-25% в сравнении с исходной Qwen2.5-VL-7B-Instruct.
+## Где используем в продукте
+* 📝 Автогенерация описаний карточек по фото
+* 🔍 Ключевые слова для поиска (извлечение признаков с изображений)
+* 🧾 OCR брендов/надписей и их нормализация
+* ⚡ «Подача объявления в один клик» по фото товара
+* 🔧 Внутренние инструменты разметки и модерации
+---
+## Quickstart (Transformers)
+Ниже — минимальный пример инференса VLM (текст+картинка).
+```python
+import torch
+from PIL import Image
+from transformers import AutoProcessor, AutoModelForImageTextToText
+from qwen_vl_utils import process_vision_info
+model_id = "AvitoTech/a-vision"
+# Модель и процессор
+model = AutoModelForImageTextToText.from_pretrained(
+    model_id,
+    torch_dtype="auto",
+    device_map="auto",
+)
+processor = AutoProcessor.from_pretrained(model_id)
+img = Image.open("assets/hoodie.jpg")  # выберите локально загруженное изображение
+messages = [
+    {
+        "role": "user",
+        "content": [
+            {
+                "type": "image",
+                "image": img,
+                "min_pixels": 4 * 28 * 28,
+                "max_pixels": 1024 * 28 * 28,
+            },
+            {
+                "type": "text",
+                "text": "Опиши изображение."
+            }
+        ],
+    }
+]
+# Подготовка входа
+chat_text = processor.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
+image_inputs, video_inputs = process_vision_info(messages)
+inputs = processor(
+    text=[chat_text],
+    images=image_inputs,
+    videos=video_inputs,
+    padding=True,
+    return_tensors="pt",
+)
+inputs = inputs.to("cuda")
+# Генерация
+generated_ids = model.generate(**inputs, max_new_tokens=256)
+generated_ids_trimmed = [
+    out_ids[len(in_ids) :] for in_ids, out_ids in zip(inputs.input_ids, generated_ids)
+]
+response = processor.batch_decode(generated_ids, skip_special_tokens=True)[0]
+print(response)
+```
+> Примечание:
+> * Для лучшей производительности на видео/мульти-картинках имеет смысл подбирать `min_pixels/max_pixels`.
+---