Sci-fi Synopsis

 AI Fine Tuning Qwen2.5

Jett recalled the moment he prepared his first dataset for fine-tuning AI. He patiently compiled 20 examples for training the neural network. He needed almost 10,000 examples. At the time, Jett was using free versions of AI from competing vendors. In about an hour, Jett generated the required number of examples and then began testing them against the "competitors." As always, nothing is done for free. Most of the examples had to be corrected, checked again, corrected again, and checked again... In the distant summer of XXXX, Jett trained an open-source AI model on a local server for the first time. The euphoria lasted for five minutes. Jett tested the trained AI model, found it extremely unsuccessful, and drew two conclusions: that good data for training AI is "more valuable than gold" and that training data is the ideal tool for "injecting" into AI. And "injections" can be made into virtually any neural network, even the most closed and isolated ones. The only difference is the length of the path.





https://huggingface.co/datasets/Oflameron/sci-fi-Synopsis


AI Fine Tuning Qwen2.5


https://docs.google.com/document/d/1U_9MFhClO-fthXlBAgoa77ESHDmh-MjCsPHadXA_Qr8/edit?usp=sharing





Комментарии

Отправить комментарий

Популярные сообщения из этого блога

JSON Data 44.5

Изображение
  AI Fine Tuning Qwen2.5 Key Features of These Examples: Standard JSON Format: Uses the instruction-input-output structure compatible with Qwen2.5-7B fine-tuning frameworks125 English Language: All content is in English as requested Mission Oflameron Context: Each example incorporates elements from your sci-fi film's lore: Virtual Consciousness concepts Jett's character and methods Transition Points technology The origin of life hypothesis Scene writing examples Diverse Task Types: Includes explanations, descriptions, dialogue writing, and narrative scenes Proper Escaping: All JSON strings are properly formatted with escaped characters where needed This listing provides 10 training examples covering various aspects of your Mission Oflameron universe, suitable for fine-tuning Qwen2.5-7B to generate content consistent with your film's concepts and characters. JSON Instruction Tuning Data 44_5 JSON_Instruction_Tuning_Data_44-5.txt Here are the updated English-language JSON exa...
Далее...

Рекомендуемый размер датасета

  AI Fine Tuning Qwen2.5 Рекомендуемый размер датасета Для fine-tuning Qwen2.5 (7B) на задаче анализа маршрутизации: Тип данных         Количество примеров Корректные таблицы 300-500 Таблицы с ошибками 300-500 Петли маршрутизации 150-250 Пограничные случаи 50-100 ИТОГО 800-1350 При этом разнообразие адресов можно обеспечить всего 50-100 уникальными IP/networks, комбинируя их в разных примерах. AI Fine Tuning Qwen2.5
Далее...

JSON Data 23

  AI Fine Tuning Qwen2.5 JSON Instruction Tuning Data 23 json {   "copyright": ["valery_shmelev", "valery_shmeleff", "oflameron"],   "series": "Mission Oflameron",   "training_examples": [     {       "scene_id": "MTLB_Briefing_001",       "scene_type": "high_level_briefing",       "location": {         "description": "A light-colored oval room with no windows, doors, or lamps. In the center is a round table and expensive office chairs.",         "atmosphere": "Secure, confidential, high-stakes meeting environment"       },       "characters": [         {           "id": "D",           "role": "Lead Scientist/Project Director",           "description": "Respectable man in expensive suit, holding tablet or compact control panel",       ...
Далее...