diff --git a/04-Model/01-Models.ipynb b/04-Model/01-Models.ipynb index 261a25702..24925a1cf 100644 --- a/04-Model/01-Models.ipynb +++ b/04-Model/01-Models.ipynb @@ -20,12 +20,12 @@ "### Table of Contents\n", "\n", "- [Overview](#overview)\n", - "- [OpenAI GPT Series](#openai---gpt-series)\n", - "- [Meta Llama Series](#meta---llama-series)\n", - "- [Anthropic Claude Series](#anthropic---claude-series)\n", - "- [Google Gemini Series](#google---gemini)\n", - "- [Mistral AI models Series](#mistral-ai-models-overview)\n", - "- [Alibaba Qwen Series](#alibaba---qwen)\n", + "- [OpenAI GPT Series](#openai-gpt-series)\n", + "- [Meta Llama Series](#meta-llama-series)\n", + "- [Anthropic Claude Series](#anthropic-claude-series)\n", + "- [Google Gemini Series](#google-gemini-series)\n", + "- [Mistral AI models Series](#mistral-ai-models-series)\n", + "- [Alibaba Qwen Series](#alibaba-qwen-series)\n", "\n", "\n", "### References\n", @@ -35,7 +35,8 @@ "- [Google’s models overview](https://ai.google.dev/gemini-api/docs/models/gemini).\n", "- [Mistral's models overview](https://mistral.ai/technology/#models).\n", "- [Alibaba Cloud’s models overview](https://mistral.ai/technology/#models).\n", - "---" + "\n", + "----" ] }, { @@ -87,10 +88,13 @@ "- Built-in safety measures\n", "- Enhanced context retention\n", "\n", - "For more detailed information, please refer to [OpenAI's official documentation](https://platform.openai.com/docs/models#models-overview).\n", - "\n", - "---\n", - "\n", + "For more detailed information, please refer to [OpenAI's official documentation](https://platform.openai.com/docs/models#models-overview)." + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ "## Meta - Llama Series\n", "\n", "Meta's Llama AI series offers open-source models that allow fine-tuning, distillation, and flexible deployment.\n", @@ -126,10 +130,19 @@ "- Pre-trained on 15 trillion tokens\n", "- Fine-tuned through Supervised Fine-tuning (SFT) and RLHF\n", "\n", - "For more detailed information, please refer to [Meta's official documentation](https://www.llama.com/).\n", - "\n", - "---\n", + " > **Supervised Fine-tuning** : Supervised fine-tuning is a process of improving an existing AI model's performance by training it with labeled data. For example, if you want to teach the model text summarization, you provide pairs of 'original text' and 'summarized text' as training data. Through this training with correct answer pairs, the model can enhance its performance on specific tasks.\n", + " >\n", + " > **Reinforcement Learning with Human Feedback (RLHF)** : RLHF is a method where AI models learn to generate better responses through human feedback. When the AI generates responses, humans evaluate them, and the model improves based on these evaluations. Just like a student improves their skills through teacher feedback, AI develops to provide more ethical and helpful responses through human feedback.\n", + " \n", + "**Use Cases** \n", "\n", + "For more detailed information, please refer to [Meta's official documentation](https://www.llama.com/).\n" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ "## Anthropic - Claude Series\n", "\n", "Claude models by Anthropic are advanced language models with cloud-based APIs for diverse NLP tasks. These models balance performance, safety, and real-time responsiveness.\n", @@ -160,10 +173,13 @@ "- Detailed technical documentation\n", "- Advanced code generation and review\n", "\n", - "For more detailed information, please refer to [Anthropic's official documentation](https://docs.anthropic.com/en/docs/intro-to-claude).\n", - "\n", - "---\n", - "\n", + "For more detailed information, please refer to [Anthropic's official documentation](https://docs.anthropic.com/en/docs/intro-to-claude).\n" + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ "## Google - Gemini\n", "\n", "Google's Gemini models prioritize efficiency and scalability, designed for a wide range of advanced applications.\n", @@ -189,10 +205,13 @@ "- Reasoning tasks for complex problem-solving \n", "- Image and text generation \n", "\n", - "For more detailed information, refer to [Google's Gemini documentation](https://ai.google.dev/gemini-api/docs/models/gemini).\n", - "\n", - "---\n", - "\n", + "For more detailed information, refer to [Google's Gemini documentation](https://ai.google.dev/gemini-api/docs/models/gemini)." + ] + }, + { + "cell_type": "markdown", + "metadata": {}, + "source": [ "## Mistral AI Models Overview\n", "\n", "Mistral AI provides commercial and open-source models for diverse NLP tasks, including specialized solutions.\n", @@ -208,10 +227,17 @@ "- Mathstral: Mathematics-focused\n", "- Codestral Mamba: 256k context for coding tasks\n", "\n", - "For more detailed information, please refer to [Mistral's official documentation](https://mistral.ai/technology/#models).\n", - "\n", - "---\n", - "\n", + "For more detailed information, please refer to [Mistral's official documentation](https://mistral.ai/technology/#models).\n" + ] + }, + { + "cell_type": "markdown", + "metadata": { + "vscode": { + "languageId": "plaintext" + } + }, + "source": [ "## Alibaba - Qwen\n", "\n", "Alibaba’s Qwen models offer open-source and commercial variants optimized for diverse industries and tasks.\n", @@ -230,19 +256,8 @@ "- Easy deployment with Alibaba Cloud’s platform\n", "- Applications in generative AI, such as writing, image generation, and audio analysis\n", "\n", - "For more detailed information, visit [Alibaba Cloud’s official Qwen page](https://mistral.ai/technology/#models).\n", - "\n" + "For more detailed information, visit [Alibaba Cloud’s official Qwen page](https://mistral.ai/technology/#models)." ] - }, - { - "cell_type": "markdown", - "metadata": {}, - "source": [] - }, - { - "cell_type": "markdown", - "metadata": {}, - "source": [] } ], "metadata": { @@ -252,4 +267,4 @@ }, "nbformat": 4, "nbformat_minor": 2 -} \ No newline at end of file +}