initial commit

Galen_AI_Demo.ipynb

@@ -0,0 +1,126 @@
+{
+  "nbformat": 4,
+  "nbformat_minor": 0,
+  "metadata": {
+    "colab": {
+      "provenance": [],
+      "gpuType": "T4"
+    },
+    "kernelspec": {
+      "name": "python3",
+      "display_name": "Python 3"
+    },
+    "language_info": {
+      "name": "python"
+    },
+    "accelerator": "GPU"
+  },
+  "cells": [
+    {
+      "cell_type": "markdown",
+      "source": [
+        "# **Galen AI Demo**"
+      ],
+      "metadata": {
+        "id": "WX4uJNZWM9T9"
+      }
+    },
+    {
+      "cell_type": "code",
+      "source": [
+        "%%capture\n",
+        "# @title Install packages { display-mode: \"form\" }\n",
+        "# @markdown Please wait for the model to be loaded, it takes minutes to load, make sure you enabled GPU in your notebook\n",
+        "\n",
+        "!pip install -q -U bitsandbytes\n",
+        "!pip install -q -U git+https://github.com/huggingface/transformers.git\n",
+        "!pip install -q -U git+https://github.com/huggingface/accelerate.git\n",
+        "!pip install -q -U einops"
+      ],
+      "metadata": {
+        "id": "o0zCCilRMuqc"
+      },
+      "execution_count": null,
+      "outputs": []
+    },
+    {
+      "cell_type": "code",
+      "source": [
+        "# @title Load the model {display-mode: \"form\"}\n",
+        "from transformers import AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig,HfArgumentParser,TrainingArguments,pipeline, logging\n",
+        "import transformers\n",
+        "import torch\n",
+        "\n",
+        "model_name = 'ahmed-ai/galen'\n",
+        "\n",
+        "tokenizer = AutoTokenizer.from_pretrained(model_name)\n",
+        "model_pipeline = pipeline(task=\"text-generation\", model=model_name, tokenizer=tokenizer)\n",
+        "\n",
+        "# bnb_config = BitsAndBytesConfig(\n",
+        "#     load_in_4bit=True,\n",
+        "#     bnb_4bit_quant_type=\"nf4\",\n",
+        "#     bnb_4bit_use_double_quant=True,\n",
+        "# )\n",
+        "# model = AutoModelForCausalLM.from_pretrained(\n",
+        "#         model_name,\n",
+        "#         load_in_4bit=True,\n",
+        "#         quantization_config=bnb_config,\n",
+        "#         torch_dtype=torch.bfloat16,\n",
+        "#         device_map=\"auto\",\n",
+        "#         trust_remote_code=True,\n",
+        "#     )"
+      ],
+      "metadata": {
+        "id": "wfV5CarBc5h_"
+      },
+      "execution_count": null,
+      "outputs": []
+    },
+    {
+      "cell_type": "code",
+      "source": [
+        "# @title Hyperparameters {display-mode: \"form\"}\n",
+        "# @markdown The hypermeters of the input\n",
+        "\n",
+        "def user_prompt(prompt):\n",
+        "    return f\"\"\"\n",
+        "  Below is an instruction that describes a task, paired with an input that provides further context. Write a response that appropriately completes the request.\n",
+        "\n",
+        "  ### Instruction:\n",
+        "  Generate an impressive qoute\n",
+        "\n",
+        "\n",
+        "  ### Input:\n",
+        "  {prompt}\n",
+        "\n",
+        "  ### Response:\n",
+        "  \"\"\"\n",
+        "\n",
+        "input = \"what is squamous cell carcinoma\" # @param {type: \"string\"}\n",
+        "max_length = 256 # @param {type:\"slider\", min:0, max:2048, step:2}\n",
+        "temperature = 0.5 # @param {type: \"slider\", min:0, max:1, step:0.1}\n",
+        "top_p = 0.5 # @param {type: \"slider\", min:0, max:1, step:0.1}"
+      ],
+      "metadata": {
+        "id": "S0ugjLBZc-1J"
+      },
+      "execution_count": null,
+      "outputs": []
+    },
+    {
+      "cell_type": "code",
+      "source": [
+        "# @title Run the model {display-mode: \"form\"}\n",
+        "\n",
+        "prompt = user_prompt(input)\n",
+        "result = model_pipeline(prompt, max_length=max_lenght, temperature=temperature, top_p=top_p)\n",
+        "print(result[0]['generated_text'][len(prompt):])"
+      ],
+      "metadata": {
+        "id": "Q9pNcQs6eyFQ"
+      },
+      "execution_count": null,
+      "outputs": []
+    }
+  ]
+}

README.md

@@ -0,0 +1,34 @@
+---
+inference: false
+datasets:
+- medalpaca/medical_meadow_medqa
+language:
+- en
+library_name: transformers
+tags:
+- biology
+- medical
+- QA
+- healthcare
+---
+# Galen
+Galen is fine-tuned from [Mistral-7B-Instruct-v0.2](https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.2), using [medical quesion answering dataset](https://huggingface.co/datasets/medalpaca/medical_meadow_medqa)
+### Galen's view about future of medicine and AI:
+![alt text](1.png "Galen's view about future of medicine and AI")
+# Get Started
+Install "accelerate" to use CUDA GPU
+```bash
+pip install accelerate
+```
+```py
+from transformers import AutoTokenizer, pipeline
+```
+```py
+tokenizer = AutoTokenizer.from_pretrained('ahmed/galen')
+model_pipeline = pipeline(task="text-generation", model='ahmed/galen', tokenizer=tokenizer, max_length=256, temperature=0.5, top_p=0.6)
+```
+```py
+result = model_pipeline('What is squamous carcinoma')
+#print the generated text
+print(result[0]['generated_text'][len(prompt):])
+```