update

Author: hamelsmu

hamel/__init__.py

@@ -1 +1 @@
-__version__ = "0.0.8"
+__version__ = "0.0.14"

nbs/00_gem.ipynb

@@ -24,7 +24,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 1,
+   "execution_count": null,
    "metadata": {},
    "outputs": [],
    "source": [
@@ -33,7 +33,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 2,
+   "execution_count": null,
    "metadata": {},
    "outputs": [],
    "source": [
@@ -52,7 +52,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 3,
+   "execution_count": null,
    "metadata": {},
    "outputs": [],
    "source": [
@@ -68,7 +68,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 4,
+   "execution_count": null,
    "metadata": {},
    "outputs": [],
    "source": [
@@ -96,7 +96,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 5,
+   "execution_count": null,
    "metadata": {},
    "outputs": [],
    "source": [
@@ -108,7 +108,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 6,
+   "execution_count": null,
    "metadata": {},
    "outputs": [],
    "source": [
@@ -127,7 +127,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 20,
+   "execution_count": null,
    "metadata": {},
    "outputs": [],
    "source": [
@@ -149,7 +149,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 21,
+   "execution_count": null,
    "metadata": {},
    "outputs": [
     {
@@ -204,7 +204,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 22,
+   "execution_count": null,
    "metadata": {},
    "outputs": [
     {
@@ -268,7 +268,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 43,
+   "execution_count": null,
    "metadata": {},
    "outputs": [],
    "source": [
@@ -306,7 +306,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 44,
+   "execution_count": null,
    "metadata": {},
    "outputs": [
     {
@@ -364,7 +364,7 @@
        ")"
       ]
      },
-     "execution_count": 44,
+     "execution_count": null,
      "metadata": {},
      "output_type": "execute_result"
     }
@@ -385,7 +385,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 48,
+   "execution_count": null,
    "metadata": {},
    "outputs": [],
    "source": [
@@ -442,7 +442,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 27,
+   "execution_count": null,
    "metadata": {},
    "outputs": [
     {
@@ -451,7 +451,7 @@
        "'Simple, clean to type,\\nIndented, logic unfolds,\\nPowerful, so swift.'"
       ]
      },
-     "execution_count": 27,
+     "execution_count": null,
      "metadata": {},
      "output_type": "execute_result"
     }
@@ -471,7 +471,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 29,
+   "execution_count": null,
    "metadata": {},
    "outputs": [
     {
@@ -480,7 +480,7 @@
        "'Late interaction overcomes single vector limits.'"
       ]
      },
-     "execution_count": 29,
+     "execution_count": null,
      "metadata": {},
      "output_type": "execute_result"
     }
@@ -501,7 +501,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 28,
+   "execution_count": null,
    "metadata": {},
    "outputs": [
     {
@@ -554,7 +554,7 @@
        "'This course aims to teach participants how to systematically evaluate LLM-powered products by mastering principles like error analysis and automated evaluators. It addresses the inherent difficulty of LLM development, characterized by \"Three Gulfs\"—comprehension of data, specification of precise instructions, and generalization to new inputs—requiring diverse skill sets. The course guides users through an iterative \"Analyze-Measure-Improve\" lifecycle, emphasizing the importance of understanding user unhappiness and effective prompting to ensure reliable, safe, and useful AI.'"
       ]
      },
-     "execution_count": 28,
+     "execution_count": null,
      "metadata": {},
      "output_type": "execute_result"
     }
@@ -575,7 +575,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 30,
+   "execution_count": null,
    "metadata": {},
    "outputs": [
     {
@@ -584,7 +584,7 @@
        "'This presentation explores \"New Frontiers in IR,\" highlighting how traditional keyword and semantic search methods fall short when users provide complex instructions or queries requiring multi-step reasoning. To address this, two models are introduced: Promptriever, a fast bi-encoder trained to follow instructions and be \"promptable\" like an LLM, and Rank1, a strong but slower cross-encoder that incorporates test-time reasoning. Evaluations show significant improvements for both, demonstrating that IR systems can achieve higher recall, greater robustness to prompts, and solve more intricate information needs by adopting these instruction-following and reasoning capabilities.'"
       ]
      },
-     "execution_count": 30,
+     "execution_count": null,
      "metadata": {},
      "output_type": "execute_result"
     }
@@ -595,7 +595,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 31,
+   "execution_count": null,
    "metadata": {},
    "outputs": [
     {
@@ -604,7 +604,7 @@
        "'This image appears to be a YouTube video thumbnail, designed to be visually engaging and informative about a technical topic, likely related to AI or data processing.\\n\\nHere\\'s a detailed description of what\\'s in the image:\\n\\n*   **Background:** A solid, dark blue or almost black background provides high contrast for the foreground elements.\\n*   **Text:**\\n    *   At the top left, in large white letters, is the question: \"Single Vector?\".\\n    *   Below it, in even larger, bright yellow capital letters, is the phrase: \"YOU\\'RE MISSING OUT\".\\n*   **Person:** On the left side of the image, positioned slightly below the \"Single Vector?\" text, is a smiling young man with light brown hair, looking directly at the viewer. He is wearing a white t-shirt. His head and upper shoulders are visible.\\n*   **Emoji:** Just above and slightly to the left of the yellow \"YOU\\'RE\" text, a yellow \"sad\" or \"disappointed\" emoji (with downturned mouth and eyebrows) is placed, partially overlapping the man\\'s head.\\n*   **Abstract Graphics (Right Side):**\\n    *   **Top Network:** A network of interconnected light blue, glowing circular nodes (resembling neurons or data points) with lines connecting them, suggesting a complex system or network structure. An arrow points downwards from this network.\\n    *   **Bottom Flow:** Another set of light blue, glowing nodes and lines leading from the bottom-left area towards a rectangular box on the right. An arrow indicates flow into this box.\\n    *   **\"RAG\" Box:** A dark blue rectangular box with rounded corners, containing the white capital letters \"RAG\". This box is the endpoint for the incoming data flow indicated by the arrows.\\n\\nThe overall composition suggests a video discussing the limitations of \"single vector\" approaches and promoting the benefits of a more complex system, possibly related to \"RAG\" (Retrieval-Augmented Generation), a technique in natural language processing. The smiling person and the \"missing out\" text combined with the sad emoji create a hook for the viewer.'"
       ]
      },
-     "execution_count": 31,
+     "execution_count": null,
      "metadata": {},
      "output_type": "execute_result"
     }
@@ -624,7 +624,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 32,
+   "execution_count": null,
    "metadata": {},
    "outputs": [
     {
@@ -633,7 +633,7 @@
        "'Hamel Husain is the **co-founder and CEO** of **Outerbounds**.\\n\\nOuterbounds is a company focused on MLOps (Machine Learning Operations) and provides an enterprise platform built around the popular open-source framework **Metaflow**, which was originally developed at Netflix.\\n\\nBefore co-founding Outerbounds in 2021, he was a Principal Machine Learning Scientist at GitHub.'"
       ]
      },
-     "execution_count": 32,
+     "execution_count": null,
      "metadata": {},
      "output_type": "execute_result"
     }
@@ -653,7 +653,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 33,
+   "execution_count": null,
    "metadata": {},
    "outputs": [
     {
@@ -662,7 +662,7 @@
        "'Hamel Husain is currently an independent consultant, assisting companies with building AI products, particularly focusing on large language models (LLMs). He is also the founder of Parlance Labs.\\n\\nPrior to his current role, he was a Staff Machine Learning Engineer at GitHub. He has over 25 years of experience in machine learning, having also worked with companies such as Airbnb and DataRobot. Additionally, he teaches a course on \"AI Evals For Engineers & PMs\".'"
       ]
      },
-     "execution_count": 33,
+     "execution_count": null,
      "metadata": {},
      "output_type": "execute_result"
     }
@@ -682,7 +682,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 34,
+   "execution_count": null,
    "metadata": {},
    "outputs": [
     {
@@ -691,7 +691,7 @@
        "'No'"
       ]
      },
-     "execution_count": 34,
+     "execution_count": null,
      "metadata": {},
      "output_type": "execute_result"
     }
@@ -703,7 +703,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 35,
+   "execution_count": null,
    "metadata": {},
    "outputs": [
     {
@@ -712,7 +712,7 @@
        "'Yes.'"
       ]
      },
-     "execution_count": 35,
+     "execution_count": null,
      "metadata": {},
      "output_type": "execute_result"
     }
@@ -723,7 +723,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 49,
+   "execution_count": null,
    "metadata": {},
    "outputs": [
     {
@@ -776,7 +776,7 @@
        "'The video and the slides do not share common content or subject matter beyond the general concept of \"testing.\"\\n\\n**Video Summary:** This video is a brief test recording featuring a man speaking his name and reciting numbers to check audio quality.\\n\\n**Slides Summary:** This presentation introduces \"Promptriever\" and \"Rank1,\" novel information retrieval models designed to follow natural language instructions and perform reasoning, proposing a shift from traditional search methods by leveraging instruction-trained retrievers and test-time computation.'"
       ]
      },
-     "execution_count": 49,
+     "execution_count": null,
      "metadata": {},
      "output_type": "execute_result"
     }
@@ -794,7 +794,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 52,
+   "execution_count": null,
    "metadata": {},
    "outputs": [],
    "source": [
@@ -812,21 +812,9 @@
  ],
  "metadata": {
   "kernelspec": {
-   "display_name": "Python 3 (ipykernel)",
+   "display_name": "python3",
    "language": "python",
    "name": "python3"
-  },
-  "language_info": {
-   "codemirror_mode": {
-    "name": "ipython",
-    "version": 3
-   },
-   "file_extension": ".py",
-   "mimetype": "text/x-python",
-   "name": "python",
-   "nbconvert_exporter": "python",
-   "pygments_lexer": "ipython3",
-   "version": "3.12.7"
   }
  },
  "nbformat": 4,