mirror of
https://github.com/hwchase17/langchain
synced 2024-10-31 15:20:26 +00:00
933 lines
27 KiB
Plaintext
933 lines
27 KiB
Plaintext
|
{
|
|||
|
"cells": [
|
|||
|
{
|
|||
|
"cell_type": "markdown",
|
|||
|
"metadata": {
|
|||
|
"id": "BYejgj8Zf-LG",
|
|||
|
"tags": []
|
|||
|
},
|
|||
|
"source": [
|
|||
|
"## Getting started with LangChain and Gemma, running locally or in the Cloud"
|
|||
|
]
|
|||
|
},
|
|||
|
{
|
|||
|
"cell_type": "markdown",
|
|||
|
"metadata": {
|
|||
|
"id": "2IxjMb9-jIJ8"
|
|||
|
},
|
|||
|
"source": [
|
|||
|
"### Installing dependencies"
|
|||
|
]
|
|||
|
},
|
|||
|
{
|
|||
|
"cell_type": "code",
|
|||
|
"execution_count": 1,
|
|||
|
"metadata": {
|
|||
|
"colab": {
|
|||
|
"base_uri": "https://localhost:8080/"
|
|||
|
},
|
|||
|
"executionInfo": {
|
|||
|
"elapsed": 9436,
|
|||
|
"status": "ok",
|
|||
|
"timestamp": 1708975187360,
|
|||
|
"user": {
|
|||
|
"displayName": "",
|
|||
|
"userId": ""
|
|||
|
},
|
|||
|
"user_tz": -60
|
|||
|
},
|
|||
|
"id": "XZaTsXfcheTF",
|
|||
|
"outputId": "eb21d603-d824-46c5-f99f-087fb2f618b1",
|
|||
|
"tags": []
|
|||
|
},
|
|||
|
"outputs": [],
|
|||
|
"source": [
|
|||
|
"!pip install --upgrade langchain langchain-google-vertexai"
|
|||
|
]
|
|||
|
},
|
|||
|
{
|
|||
|
"cell_type": "markdown",
|
|||
|
"metadata": {
|
|||
|
"id": "IXmAujvC3Kwp"
|
|||
|
},
|
|||
|
"source": [
|
|||
|
"### Running the model"
|
|||
|
]
|
|||
|
},
|
|||
|
{
|
|||
|
"cell_type": "markdown",
|
|||
|
"metadata": {
|
|||
|
"id": "CI8Elyc5gBQF"
|
|||
|
},
|
|||
|
"source": [
|
|||
|
"Go to the VertexAI Model Garden on Google Cloud [console](https://pantheon.corp.google.com/vertex-ai/publishers/google/model-garden/335), and deploy the desired version of Gemma to VertexAI. It will take a few minutes, and after the endpoint it ready, you need to copy its number."
|
|||
|
]
|
|||
|
},
|
|||
|
{
|
|||
|
"cell_type": "code",
|
|||
|
"execution_count": 1,
|
|||
|
"metadata": {
|
|||
|
"id": "gv1j8FrVftsC"
|
|||
|
},
|
|||
|
"outputs": [],
|
|||
|
"source": [
|
|||
|
"# @title Basic parameters\n",
|
|||
|
"project: str = \"PUT_YOUR_PROJECT_ID_HERE\" # @param {type:\"string\"}\n",
|
|||
|
"endpoint_id: str = \"PUT_YOUR_ENDPOINT_ID_HERE\" # @param {type:\"string\"}\n",
|
|||
|
"location: str = \"PUT_YOUR_ENDPOINT_LOCAtION_HERE\" # @param {type:\"string\"}"
|
|||
|
]
|
|||
|
},
|
|||
|
{
|
|||
|
"cell_type": "code",
|
|||
|
"execution_count": 3,
|
|||
|
"metadata": {
|
|||
|
"executionInfo": {
|
|||
|
"elapsed": 3,
|
|||
|
"status": "ok",
|
|||
|
"timestamp": 1708975440503,
|
|||
|
"user": {
|
|||
|
"displayName": "",
|
|||
|
"userId": ""
|
|||
|
},
|
|||
|
"user_tz": -60
|
|||
|
},
|
|||
|
"id": "bhIHsFGYjtFt",
|
|||
|
"tags": []
|
|||
|
},
|
|||
|
"outputs": [
|
|||
|
{
|
|||
|
"name": "stderr",
|
|||
|
"output_type": "stream",
|
|||
|
"text": [
|
|||
|
"2024-02-27 17:15:10.457149: I tensorflow/core/util/port.cc:113] oneDNN custom operations are on. You may see slightly different numerical results due to floating-point round-off errors from different computation orders. To turn them off, set the environment variable `TF_ENABLE_ONEDNN_OPTS=0`.\n",
|
|||
|
"2024-02-27 17:15:10.508925: E external/local_xla/xla/stream_executor/cuda/cuda_dnn.cc:9261] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered\n",
|
|||
|
"2024-02-27 17:15:10.508957: E external/local_xla/xla/stream_executor/cuda/cuda_fft.cc:607] Unable to register cuFFT factory: Attempting to register factory for plugin cuFFT when one has already been registered\n",
|
|||
|
"2024-02-27 17:15:10.510289: E external/local_xla/xla/stream_executor/cuda/cuda_blas.cc:1515] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered\n",
|
|||
|
"2024-02-27 17:15:10.518898: I tensorflow/core/platform/cpu_feature_guard.cc:182] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations.\n",
|
|||
|
"To enable the following instructions: AVX2 AVX512F AVX512_VNNI FMA, in other operations, rebuild TensorFlow with the appropriate compiler flags.\n"
|
|||
|
]
|
|||
|
}
|
|||
|
],
|
|||
|
"source": [
|
|||
|
"from langchain_google_vertexai import (\n",
|
|||
|
" GemmaChatVertexAIModelGarden,\n",
|
|||
|
" GemmaVertexAIModelGarden,\n",
|
|||
|
")"
|
|||
|
]
|
|||
|
},
|
|||
|
{
|
|||
|
"cell_type": "code",
|
|||
|
"execution_count": 4,
|
|||
|
"metadata": {
|
|||
|
"executionInfo": {
|
|||
|
"elapsed": 351,
|
|||
|
"status": "ok",
|
|||
|
"timestamp": 1708975440852,
|
|||
|
"user": {
|
|||
|
"displayName": "",
|
|||
|
"userId": ""
|
|||
|
},
|
|||
|
"user_tz": -60
|
|||
|
},
|
|||
|
"id": "WJv-UVWwh0lk",
|
|||
|
"tags": []
|
|||
|
},
|
|||
|
"outputs": [],
|
|||
|
"source": [
|
|||
|
"llm = GemmaVertexAIModelGarden(\n",
|
|||
|
" endpoint_id=endpoint_id,\n",
|
|||
|
" project=project,\n",
|
|||
|
" location=location,\n",
|
|||
|
")"
|
|||
|
]
|
|||
|
},
|
|||
|
{
|
|||
|
"cell_type": "code",
|
|||
|
"execution_count": 5,
|
|||
|
"metadata": {
|
|||
|
"colab": {
|
|||
|
"base_uri": "https://localhost:8080/"
|
|||
|
},
|
|||
|
"executionInfo": {
|
|||
|
"elapsed": 714,
|
|||
|
"status": "ok",
|
|||
|
"timestamp": 1708975441564,
|
|||
|
"user": {
|
|||
|
"displayName": "",
|
|||
|
"userId": ""
|
|||
|
},
|
|||
|
"user_tz": -60
|
|||
|
},
|
|||
|
"id": "6kM7cEFdiN9h",
|
|||
|
"outputId": "fb420c56-5614-4745-cda8-0ee450a3e539",
|
|||
|
"tags": []
|
|||
|
},
|
|||
|
"outputs": [
|
|||
|
{
|
|||
|
"name": "stdout",
|
|||
|
"output_type": "stream",
|
|||
|
"text": [
|
|||
|
"Prompt:\n",
|
|||
|
"What is the meaning of life?\n",
|
|||
|
"Output:\n",
|
|||
|
" Who am I? Why do I exist? These are questions I have struggled with\n"
|
|||
|
]
|
|||
|
}
|
|||
|
],
|
|||
|
"source": [
|
|||
|
"output = llm.invoke(\"What is the meaning of life?\")\n",
|
|||
|
"print(output)"
|
|||
|
]
|
|||
|
},
|
|||
|
{
|
|||
|
"cell_type": "markdown",
|
|||
|
"metadata": {
|
|||
|
"id": "zzep9nfmuUcO"
|
|||
|
},
|
|||
|
"source": [
|
|||
|
"We can also use Gemma as a multi-turn chat model:"
|
|||
|
]
|
|||
|
},
|
|||
|
{
|
|||
|
"cell_type": "code",
|
|||
|
"execution_count": 7,
|
|||
|
"metadata": {
|
|||
|
"colab": {
|
|||
|
"base_uri": "https://localhost:8080/"
|
|||
|
},
|
|||
|
"executionInfo": {
|
|||
|
"elapsed": 964,
|
|||
|
"status": "ok",
|
|||
|
"timestamp": 1708976298189,
|
|||
|
"user": {
|
|||
|
"displayName": "",
|
|||
|
"userId": ""
|
|||
|
},
|
|||
|
"user_tz": -60
|
|||
|
},
|
|||
|
"id": "8tPHoM5XiZOl",
|
|||
|
"outputId": "7b8fb652-9aed-47b0-c096-aa1abfc3a2a9",
|
|||
|
"tags": []
|
|||
|
},
|
|||
|
"outputs": [
|
|||
|
{
|
|||
|
"name": "stdout",
|
|||
|
"output_type": "stream",
|
|||
|
"text": [
|
|||
|
"content='Prompt:\\n<start_of_turn>user\\nHow much is 2+2?<end_of_turn>\\n<start_of_turn>model\\nOutput:\\n8-years old.<end_of_turn>\\n\\n<start_of'\n",
|
|||
|
"content='Prompt:\\n<start_of_turn>user\\nHow much is 2+2?<end_of_turn>\\n<start_of_turn>model\\nPrompt:\\n<start_of_turn>user\\nHow much is 2+2?<end_of_turn>\\n<start_of_turn>model\\nOutput:\\n8-years old.<end_of_turn>\\n\\n<start_of<end_of_turn>\\n<start_of_turn>user\\nHow much is 3+3?<end_of_turn>\\n<start_of_turn>model\\nOutput:\\nOutput:\\n3-years old.<end_of_turn>\\n\\n<'\n"
|
|||
|
]
|
|||
|
}
|
|||
|
],
|
|||
|
"source": [
|
|||
|
"from langchain_core.messages import HumanMessage\n",
|
|||
|
"\n",
|
|||
|
"llm = GemmaChatVertexAIModelGarden(\n",
|
|||
|
" endpoint_id=endpoint_id,\n",
|
|||
|
" project=project,\n",
|
|||
|
" location=location,\n",
|
|||
|
")\n",
|
|||
|
"\n",
|
|||
|
"message1 = HumanMessage(content=\"How much is 2+2?\")\n",
|
|||
|
"answer1 = llm.invoke([message1])\n",
|
|||
|
"print(answer1)\n",
|
|||
|
"\n",
|
|||
|
"message2 = HumanMessage(content=\"How much is 3+3?\")\n",
|
|||
|
"answer2 = llm.invoke([message1, answer1, message2])\n",
|
|||
|
"\n",
|
|||
|
"print(answer2)"
|
|||
|
]
|
|||
|
},
|
|||
|
{
|
|||
|
"cell_type": "markdown",
|
|||
|
"metadata": {},
|
|||
|
"source": [
|
|||
|
"You can post-process response to avoid repetitions:"
|
|||
|
]
|
|||
|
},
|
|||
|
{
|
|||
|
"cell_type": "code",
|
|||
|
"execution_count": 8,
|
|||
|
"metadata": {
|
|||
|
"tags": []
|
|||
|
},
|
|||
|
"outputs": [
|
|||
|
{
|
|||
|
"name": "stdout",
|
|||
|
"output_type": "stream",
|
|||
|
"text": [
|
|||
|
"content='Output:\\n<<humming>>: 2+2 = 4.\\n<end'\n",
|
|||
|
"content='Output:\\nOutput:\\n<<humming>>: 3+3 = 6.'\n"
|
|||
|
]
|
|||
|
}
|
|||
|
],
|
|||
|
"source": [
|
|||
|
"answer1 = llm.invoke([message1], parse_response=True)\n",
|
|||
|
"print(answer1)\n",
|
|||
|
"\n",
|
|||
|
"answer2 = llm.invoke([message1, answer1, message2], parse_response=True)\n",
|
|||
|
"\n",
|
|||
|
"print(answer2)"
|
|||
|
]
|
|||
|
},
|
|||
|
{
|
|||
|
"cell_type": "markdown",
|
|||
|
"metadata": {
|
|||
|
"id": "VEfjqo7fjARR"
|
|||
|
},
|
|||
|
"source": [
|
|||
|
"## Running Gemma locally from Kaggle"
|
|||
|
]
|
|||
|
},
|
|||
|
{
|
|||
|
"cell_type": "markdown",
|
|||
|
"metadata": {
|
|||
|
"id": "gVW8QDzHu7TA"
|
|||
|
},
|
|||
|
"source": [
|
|||
|
"In order to run Gemma locally, you can download it from Kaggle first. In order to do this, you'll need to login into the Kaggle platform, create a API key and download a `kaggle.json` Read more about Kaggle auth [here](https://www.kaggle.com/docs/api)."
|
|||
|
]
|
|||
|
},
|
|||
|
{
|
|||
|
"cell_type": "markdown",
|
|||
|
"metadata": {
|
|||
|
"id": "S1EsXQ3XvZkQ"
|
|||
|
},
|
|||
|
"source": [
|
|||
|
"### Installation"
|
|||
|
]
|
|||
|
},
|
|||
|
{
|
|||
|
"cell_type": "code",
|
|||
|
"execution_count": 7,
|
|||
|
"metadata": {
|
|||
|
"executionInfo": {
|
|||
|
"elapsed": 335,
|
|||
|
"status": "ok",
|
|||
|
"timestamp": 1708976305471,
|
|||
|
"user": {
|
|||
|
"displayName": "",
|
|||
|
"userId": ""
|
|||
|
},
|
|||
|
"user_tz": -60
|
|||
|
},
|
|||
|
"id": "p8SMwpKRvbef",
|
|||
|
"tags": []
|
|||
|
},
|
|||
|
"outputs": [
|
|||
|
{
|
|||
|
"name": "stderr",
|
|||
|
"output_type": "stream",
|
|||
|
"text": [
|
|||
|
"/opt/conda/lib/python3.10/pty.py:89: RuntimeWarning: os.fork() was called. os.fork() is incompatible with multithreaded code, and JAX is multithreaded, so this will likely lead to a deadlock.\n",
|
|||
|
" pid, fd = os.forkpty()\n"
|
|||
|
]
|
|||
|
}
|
|||
|
],
|
|||
|
"source": [
|
|||
|
"!mkdir -p ~/.kaggle && cp kaggle.json ~/.kaggle/kaggle.json"
|
|||
|
]
|
|||
|
},
|
|||
|
{
|
|||
|
"cell_type": "code",
|
|||
|
"execution_count": 11,
|
|||
|
"metadata": {
|
|||
|
"executionInfo": {
|
|||
|
"elapsed": 7802,
|
|||
|
"status": "ok",
|
|||
|
"timestamp": 1708976363010,
|
|||
|
"user": {
|
|||
|
"displayName": "",
|
|||
|
"userId": ""
|
|||
|
},
|
|||
|
"user_tz": -60
|
|||
|
},
|
|||
|
"id": "Yr679aePv9Fq",
|
|||
|
"tags": []
|
|||
|
},
|
|||
|
"outputs": [
|
|||
|
{
|
|||
|
"name": "stderr",
|
|||
|
"output_type": "stream",
|
|||
|
"text": [
|
|||
|
"/opt/conda/lib/python3.10/pty.py:89: RuntimeWarning: os.fork() was called. os.fork() is incompatible with multithreaded code, and JAX is multithreaded, so this will likely lead to a deadlock.\n",
|
|||
|
" pid, fd = os.forkpty()\n"
|
|||
|
]
|
|||
|
},
|
|||
|
{
|
|||
|
"name": "stdout",
|
|||
|
"output_type": "stream",
|
|||
|
"text": [
|
|||
|
"\u001b[31mERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.\n",
|
|||
|
"tensorstore 0.1.54 requires ml-dtypes>=0.3.1, but you have ml-dtypes 0.2.0 which is incompatible.\u001b[0m\u001b[31m\n",
|
|||
|
"\u001b[0m"
|
|||
|
]
|
|||
|
}
|
|||
|
],
|
|||
|
"source": [
|
|||
|
"!pip install keras>=3 keras_nlp"
|
|||
|
]
|
|||
|
},
|
|||
|
{
|
|||
|
"cell_type": "markdown",
|
|||
|
"metadata": {
|
|||
|
"id": "E9zn8nYpv3QZ"
|
|||
|
},
|
|||
|
"source": [
|
|||
|
"### Usage"
|
|||
|
]
|
|||
|
},
|
|||
|
{
|
|||
|
"cell_type": "code",
|
|||
|
"execution_count": 1,
|
|||
|
"metadata": {
|
|||
|
"executionInfo": {
|
|||
|
"elapsed": 8536,
|
|||
|
"status": "ok",
|
|||
|
"timestamp": 1708976601206,
|
|||
|
"user": {
|
|||
|
"displayName": "",
|
|||
|
"userId": ""
|
|||
|
},
|
|||
|
"user_tz": -60
|
|||
|
},
|
|||
|
"id": "0LFRmY8TjCkI",
|
|||
|
"tags": []
|
|||
|
},
|
|||
|
"outputs": [
|
|||
|
{
|
|||
|
"name": "stderr",
|
|||
|
"output_type": "stream",
|
|||
|
"text": [
|
|||
|
"2024-02-27 16:38:40.797559: I tensorflow/core/util/port.cc:113] oneDNN custom operations are on. You may see slightly different numerical results due to floating-point round-off errors from different computation orders. To turn them off, set the environment variable `TF_ENABLE_ONEDNN_OPTS=0`.\n",
|
|||
|
"2024-02-27 16:38:40.848444: E external/local_xla/xla/stream_executor/cuda/cuda_dnn.cc:9261] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered\n",
|
|||
|
"2024-02-27 16:38:40.848478: E external/local_xla/xla/stream_executor/cuda/cuda_fft.cc:607] Unable to register cuFFT factory: Attempting to register factory for plugin cuFFT when one has already been registered\n",
|
|||
|
"2024-02-27 16:38:40.849728: E external/local_xla/xla/stream_executor/cuda/cuda_blas.cc:1515] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered\n",
|
|||
|
"2024-02-27 16:38:40.857936: I tensorflow/core/platform/cpu_feature_guard.cc:182] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations.\n",
|
|||
|
"To enable the following instructions: AVX2 AVX512F AVX512_VNNI FMA, in other operations, rebuild TensorFlow with the appropriate compiler flags.\n"
|
|||
|
]
|
|||
|
}
|
|||
|
],
|
|||
|
"source": [
|
|||
|
"from langchain_google_vertexai import GemmaLocalKaggle"
|
|||
|
]
|
|||
|
},
|
|||
|
{
|
|||
|
"cell_type": "markdown",
|
|||
|
"metadata": {
|
|||
|
"id": "v-o7oXVavdMQ"
|
|||
|
},
|
|||
|
"source": [
|
|||
|
"You can specify the keras backend (by default it's `tensorflow`, but you can change it be `jax` or `torch`)."
|
|||
|
]
|
|||
|
},
|
|||
|
{
|
|||
|
"cell_type": "code",
|
|||
|
"execution_count": 2,
|
|||
|
"metadata": {
|
|||
|
"executionInfo": {
|
|||
|
"elapsed": 9,
|
|||
|
"status": "ok",
|
|||
|
"timestamp": 1708976601206,
|
|||
|
"user": {
|
|||
|
"displayName": "",
|
|||
|
"userId": ""
|
|||
|
},
|
|||
|
"user_tz": -60
|
|||
|
},
|
|||
|
"id": "vvTUH8DNj5SF",
|
|||
|
"tags": []
|
|||
|
},
|
|||
|
"outputs": [],
|
|||
|
"source": [
|
|||
|
"# @title Basic parameters\n",
|
|||
|
"keras_backend: str = \"jax\" # @param {type:\"string\"}\n",
|
|||
|
"model_name: str = \"gemma_2b_en\" # @param {type:\"string\"}"
|
|||
|
]
|
|||
|
},
|
|||
|
{
|
|||
|
"cell_type": "code",
|
|||
|
"execution_count": 3,
|
|||
|
"metadata": {
|
|||
|
"executionInfo": {
|
|||
|
"elapsed": 40836,
|
|||
|
"status": "ok",
|
|||
|
"timestamp": 1708976761257,
|
|||
|
"user": {
|
|||
|
"displayName": "",
|
|||
|
"userId": ""
|
|||
|
},
|
|||
|
"user_tz": -60
|
|||
|
},
|
|||
|
"id": "YOmrqxo5kHXK",
|
|||
|
"tags": []
|
|||
|
},
|
|||
|
"outputs": [
|
|||
|
{
|
|||
|
"name": "stderr",
|
|||
|
"output_type": "stream",
|
|||
|
"text": [
|
|||
|
"2024-02-27 16:23:14.661164: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1929] Created device /job:localhost/replica:0/task:0/device:GPU:0 with 20549 MB memory: -> device: 0, name: NVIDIA L4, pci bus id: 0000:00:03.0, compute capability: 8.9\n",
|
|||
|
"normalizer.cc(51) LOG(INFO) precompiled_charsmap is empty. use identity normalization.\n"
|
|||
|
]
|
|||
|
}
|
|||
|
],
|
|||
|
"source": [
|
|||
|
"llm = GemmaLocalKaggle(model_name=model_name, keras_backend=keras_backend)"
|
|||
|
]
|
|||
|
},
|
|||
|
{
|
|||
|
"cell_type": "code",
|
|||
|
"execution_count": 7,
|
|||
|
"metadata": {
|
|||
|
"id": "Zu6yPDUgkQtQ",
|
|||
|
"tags": []
|
|||
|
},
|
|||
|
"outputs": [
|
|||
|
{
|
|||
|
"name": "stderr",
|
|||
|
"output_type": "stream",
|
|||
|
"text": [
|
|||
|
"W0000 00:00:1709051129.518076 774855 graph_launch.cc:671] Fallback to op-by-op mode because memset node breaks graph update\n"
|
|||
|
]
|
|||
|
},
|
|||
|
{
|
|||
|
"name": "stdout",
|
|||
|
"output_type": "stream",
|
|||
|
"text": [
|
|||
|
"What is the meaning of life?\n",
|
|||
|
"\n",
|
|||
|
"The question is one of the most important questions in the world.\n",
|
|||
|
"\n",
|
|||
|
"It’s the question that has\n"
|
|||
|
]
|
|||
|
}
|
|||
|
],
|
|||
|
"source": [
|
|||
|
"output = llm.invoke(\"What is the meaning of life?\", max_tokens=30)\n",
|
|||
|
"print(output)"
|
|||
|
]
|
|||
|
},
|
|||
|
{
|
|||
|
"cell_type": "markdown",
|
|||
|
"metadata": {},
|
|||
|
"source": [
|
|||
|
"### ChatModel"
|
|||
|
]
|
|||
|
},
|
|||
|
{
|
|||
|
"cell_type": "markdown",
|
|||
|
"metadata": {
|
|||
|
"id": "MSctpRE4u43N"
|
|||
|
},
|
|||
|
"source": [
|
|||
|
"Same as above, using Gemma locally as a multi-turn chat model. You might need to re-start the notebook and clean your GPU memory in order to avoid OOM errors:"
|
|||
|
]
|
|||
|
},
|
|||
|
{
|
|||
|
"cell_type": "code",
|
|||
|
"execution_count": 1,
|
|||
|
"metadata": {
|
|||
|
"tags": []
|
|||
|
},
|
|||
|
"outputs": [
|
|||
|
{
|
|||
|
"name": "stderr",
|
|||
|
"output_type": "stream",
|
|||
|
"text": [
|
|||
|
"2024-02-27 16:58:22.331067: I tensorflow/core/util/port.cc:113] oneDNN custom operations are on. You may see slightly different numerical results due to floating-point round-off errors from different computation orders. To turn them off, set the environment variable `TF_ENABLE_ONEDNN_OPTS=0`.\n",
|
|||
|
"2024-02-27 16:58:22.382948: E external/local_xla/xla/stream_executor/cuda/cuda_dnn.cc:9261] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered\n",
|
|||
|
"2024-02-27 16:58:22.382978: E external/local_xla/xla/stream_executor/cuda/cuda_fft.cc:607] Unable to register cuFFT factory: Attempting to register factory for plugin cuFFT when one has already been registered\n",
|
|||
|
"2024-02-27 16:58:22.384312: E external/local_xla/xla/stream_executor/cuda/cuda_blas.cc:1515] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered\n",
|
|||
|
"2024-02-27 16:58:22.392767: I tensorflow/core/platform/cpu_feature_guard.cc:182] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations.\n",
|
|||
|
"To enable the following instructions: AVX2 AVX512F AVX512_VNNI FMA, in other operations, rebuild TensorFlow with the appropriate compiler flags.\n"
|
|||
|
]
|
|||
|
}
|
|||
|
],
|
|||
|
"source": [
|
|||
|
"from langchain_google_vertexai import GemmaChatLocalKaggle"
|
|||
|
]
|
|||
|
},
|
|||
|
{
|
|||
|
"cell_type": "code",
|
|||
|
"execution_count": 2,
|
|||
|
"metadata": {
|
|||
|
"tags": []
|
|||
|
},
|
|||
|
"outputs": [],
|
|||
|
"source": [
|
|||
|
"# @title Basic parameters\n",
|
|||
|
"keras_backend: str = \"jax\" # @param {type:\"string\"}\n",
|
|||
|
"model_name: str = \"gemma_2b_en\" # @param {type:\"string\"}"
|
|||
|
]
|
|||
|
},
|
|||
|
{
|
|||
|
"cell_type": "code",
|
|||
|
"execution_count": 3,
|
|||
|
"metadata": {
|
|||
|
"tags": []
|
|||
|
},
|
|||
|
"outputs": [
|
|||
|
{
|
|||
|
"name": "stderr",
|
|||
|
"output_type": "stream",
|
|||
|
"text": [
|
|||
|
"2024-02-27 16:58:29.001922: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1929] Created device /job:localhost/replica:0/task:0/device:GPU:0 with 20549 MB memory: -> device: 0, name: NVIDIA L4, pci bus id: 0000:00:03.0, compute capability: 8.9\n",
|
|||
|
"normalizer.cc(51) LOG(INFO) precompiled_charsmap is empty. use identity normalization.\n"
|
|||
|
]
|
|||
|
}
|
|||
|
],
|
|||
|
"source": [
|
|||
|
"llm = GemmaChatLocalKaggle(model_name=model_name, keras_backend=keras_backend)"
|
|||
|
]
|
|||
|
},
|
|||
|
{
|
|||
|
"cell_type": "code",
|
|||
|
"execution_count": 4,
|
|||
|
"metadata": {
|
|||
|
"executionInfo": {
|
|||
|
"elapsed": 3,
|
|||
|
"status": "aborted",
|
|||
|
"timestamp": 1708976382957,
|
|||
|
"user": {
|
|||
|
"displayName": "",
|
|||
|
"userId": ""
|
|||
|
},
|
|||
|
"user_tz": -60
|
|||
|
},
|
|||
|
"id": "JrJmvZqwwLqj"
|
|||
|
},
|
|||
|
"outputs": [
|
|||
|
{
|
|||
|
"name": "stderr",
|
|||
|
"output_type": "stream",
|
|||
|
"text": [
|
|||
|
"2024-02-27 16:58:49.848412: I external/local_xla/xla/service/service.cc:168] XLA service 0x55adc0cf2c10 initialized for platform CUDA (this does not guarantee that XLA will be used). Devices:\n",
|
|||
|
"2024-02-27 16:58:49.848458: I external/local_xla/xla/service/service.cc:176] StreamExecutor device (0): NVIDIA L4, Compute Capability 8.9\n",
|
|||
|
"2024-02-27 16:58:50.116614: I tensorflow/compiler/mlir/tensorflow/utils/dump_mlir_util.cc:269] disabling MLIR crash reproducer, set env var `MLIR_CRASH_REPRODUCER_DIRECTORY` to enable.\n",
|
|||
|
"2024-02-27 16:58:54.389324: I external/local_xla/xla/stream_executor/cuda/cuda_dnn.cc:454] Loaded cuDNN version 8900\n",
|
|||
|
"WARNING: All log messages before absl::InitializeLog() is called are written to STDERR\n",
|
|||
|
"I0000 00:00:1709053145.225207 784891 device_compiler.h:186] Compiled cluster using XLA! This line is logged at most once for the lifetime of the process.\n",
|
|||
|
"W0000 00:00:1709053145.284227 784891 graph_launch.cc:671] Fallback to op-by-op mode because memset node breaks graph update\n"
|
|||
|
]
|
|||
|
},
|
|||
|
{
|
|||
|
"name": "stdout",
|
|||
|
"output_type": "stream",
|
|||
|
"text": [
|
|||
|
"content=\"<start_of_turn>user\\nHi! Who are you?<end_of_turn>\\n<start_of_turn>model\\nI'm a model.\\n Tampoco\\nI'm a model.\"\n"
|
|||
|
]
|
|||
|
}
|
|||
|
],
|
|||
|
"source": [
|
|||
|
"from langchain_core.messages import HumanMessage\n",
|
|||
|
"\n",
|
|||
|
"message1 = HumanMessage(content=\"Hi! Who are you?\")\n",
|
|||
|
"answer1 = llm.invoke([message1], max_tokens=30)\n",
|
|||
|
"print(answer1)"
|
|||
|
]
|
|||
|
},
|
|||
|
{
|
|||
|
"cell_type": "code",
|
|||
|
"execution_count": 5,
|
|||
|
"metadata": {
|
|||
|
"tags": []
|
|||
|
},
|
|||
|
"outputs": [
|
|||
|
{
|
|||
|
"name": "stdout",
|
|||
|
"output_type": "stream",
|
|||
|
"text": [
|
|||
|
"content=\"<start_of_turn>user\\nHi! Who are you?<end_of_turn>\\n<start_of_turn>model\\n<start_of_turn>user\\nHi! Who are you?<end_of_turn>\\n<start_of_turn>model\\nI'm a model.\\n Tampoco\\nI'm a model.<end_of_turn>\\n<start_of_turn>user\\nWhat can you help me with?<end_of_turn>\\n<start_of_turn>model\"\n"
|
|||
|
]
|
|||
|
}
|
|||
|
],
|
|||
|
"source": [
|
|||
|
"message2 = HumanMessage(content=\"What can you help me with?\")\n",
|
|||
|
"answer2 = llm.invoke([message1, answer1, message2], max_tokens=60)\n",
|
|||
|
"\n",
|
|||
|
"print(answer2)"
|
|||
|
]
|
|||
|
},
|
|||
|
{
|
|||
|
"cell_type": "markdown",
|
|||
|
"metadata": {},
|
|||
|
"source": [
|
|||
|
"You can post-process the response if you want to avoid multi-turn statements:"
|
|||
|
]
|
|||
|
},
|
|||
|
{
|
|||
|
"cell_type": "code",
|
|||
|
"execution_count": 7,
|
|||
|
"metadata": {
|
|||
|
"tags": []
|
|||
|
},
|
|||
|
"outputs": [
|
|||
|
{
|
|||
|
"name": "stdout",
|
|||
|
"output_type": "stream",
|
|||
|
"text": [
|
|||
|
"content=\"I'm a model.\\n Tampoco\\nI'm a model.\"\n",
|
|||
|
"content='I can help you with your modeling.\\n Tampoco\\nI can'\n"
|
|||
|
]
|
|||
|
}
|
|||
|
],
|
|||
|
"source": [
|
|||
|
"answer1 = llm.invoke([message1], max_tokens=30, parse_response=True)\n",
|
|||
|
"print(answer1)\n",
|
|||
|
"\n",
|
|||
|
"answer2 = llm.invoke([message1, answer1, message2], max_tokens=60, parse_response=True)\n",
|
|||
|
"print(answer2)"
|
|||
|
]
|
|||
|
},
|
|||
|
{
|
|||
|
"cell_type": "markdown",
|
|||
|
"metadata": {
|
|||
|
"id": "EiZnztso7hyF"
|
|||
|
},
|
|||
|
"source": [
|
|||
|
"## Running Gemma locally from HuggingFace"
|
|||
|
]
|
|||
|
},
|
|||
|
{
|
|||
|
"cell_type": "code",
|
|||
|
"execution_count": 1,
|
|||
|
"metadata": {
|
|||
|
"id": "qqAqsz5R7nKf",
|
|||
|
"tags": []
|
|||
|
},
|
|||
|
"outputs": [
|
|||
|
{
|
|||
|
"name": "stderr",
|
|||
|
"output_type": "stream",
|
|||
|
"text": [
|
|||
|
"2024-02-27 17:02:21.832409: I tensorflow/core/util/port.cc:113] oneDNN custom operations are on. You may see slightly different numerical results due to floating-point round-off errors from different computation orders. To turn them off, set the environment variable `TF_ENABLE_ONEDNN_OPTS=0`.\n",
|
|||
|
"2024-02-27 17:02:21.883625: E external/local_xla/xla/stream_executor/cuda/cuda_dnn.cc:9261] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered\n",
|
|||
|
"2024-02-27 17:02:21.883656: E external/local_xla/xla/stream_executor/cuda/cuda_fft.cc:607] Unable to register cuFFT factory: Attempting to register factory for plugin cuFFT when one has already been registered\n",
|
|||
|
"2024-02-27 17:02:21.884987: E external/local_xla/xla/stream_executor/cuda/cuda_blas.cc:1515] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered\n",
|
|||
|
"2024-02-27 17:02:21.893340: I tensorflow/core/platform/cpu_feature_guard.cc:182] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations.\n",
|
|||
|
"To enable the following instructions: AVX2 AVX512F AVX512_VNNI FMA, in other operations, rebuild TensorFlow with the appropriate compiler flags.\n"
|
|||
|
]
|
|||
|
}
|
|||
|
],
|
|||
|
"source": [
|
|||
|
"from langchain_google_vertexai import GemmaChatLocalHF, GemmaLocalHF"
|
|||
|
]
|
|||
|
},
|
|||
|
{
|
|||
|
"cell_type": "code",
|
|||
|
"execution_count": 2,
|
|||
|
"metadata": {
|
|||
|
"id": "tsyntzI08cOr",
|
|||
|
"tags": []
|
|||
|
},
|
|||
|
"outputs": [],
|
|||
|
"source": [
|
|||
|
"# @title Basic parameters\n",
|
|||
|
"hf_access_token: str = \"PUT_YOUR_TOKEN_HERE\" # @param {type:\"string\"}\n",
|
|||
|
"model_name: str = \"google/gemma-2b\" # @param {type:\"string\"}"
|
|||
|
]
|
|||
|
},
|
|||
|
{
|
|||
|
"cell_type": "code",
|
|||
|
"execution_count": 4,
|
|||
|
"metadata": {
|
|||
|
"id": "JWrqEkOo8sm9",
|
|||
|
"tags": []
|
|||
|
},
|
|||
|
"outputs": [
|
|||
|
{
|
|||
|
"data": {
|
|||
|
"application/vnd.jupyter.widget-view+json": {
|
|||
|
"model_id": "a0d6de5542254ed1b6d3ba65465e050e",
|
|||
|
"version_major": 2,
|
|||
|
"version_minor": 0
|
|||
|
},
|
|||
|
"text/plain": [
|
|||
|
"Loading checkpoint shards: 0%| | 0/2 [00:00<?, ?it/s]"
|
|||
|
]
|
|||
|
},
|
|||
|
"metadata": {},
|
|||
|
"output_type": "display_data"
|
|||
|
}
|
|||
|
],
|
|||
|
"source": [
|
|||
|
"llm = GemmaLocalHF(model_name=\"google/gemma-2b\", hf_access_token=hf_access_token)"
|
|||
|
]
|
|||
|
},
|
|||
|
{
|
|||
|
"cell_type": "code",
|
|||
|
"execution_count": 6,
|
|||
|
"metadata": {
|
|||
|
"id": "VX96Jf4Y84k-",
|
|||
|
"tags": []
|
|||
|
},
|
|||
|
"outputs": [
|
|||
|
{
|
|||
|
"name": "stdout",
|
|||
|
"output_type": "stream",
|
|||
|
"text": [
|
|||
|
"What is the meaning of life?\n",
|
|||
|
"\n",
|
|||
|
"The question is one of the most important questions in the world.\n",
|
|||
|
"\n",
|
|||
|
"It’s the question that has been asked by philosophers, theologians, and scientists for centuries.\n",
|
|||
|
"\n",
|
|||
|
"And it’s the question that\n"
|
|||
|
]
|
|||
|
}
|
|||
|
],
|
|||
|
"source": [
|
|||
|
"output = llm.invoke(\"What is the meaning of life?\", max_tokens=50)\n",
|
|||
|
"print(output)"
|
|||
|
]
|
|||
|
},
|
|||
|
{
|
|||
|
"cell_type": "markdown",
|
|||
|
"metadata": {},
|
|||
|
"source": [
|
|||
|
"Same as above, using Gemma locally as a multi-turn chat model. You might need to re-start the notebook and clean your GPU memory in order to avoid OOM errors:"
|
|||
|
]
|
|||
|
},
|
|||
|
{
|
|||
|
"cell_type": "code",
|
|||
|
"execution_count": 3,
|
|||
|
"metadata": {
|
|||
|
"id": "9x-jmEBg9Mk1"
|
|||
|
},
|
|||
|
"outputs": [
|
|||
|
{
|
|||
|
"data": {
|
|||
|
"application/vnd.jupyter.widget-view+json": {
|
|||
|
"model_id": "c9a0b8e161d74a6faca83b1be96dee27",
|
|||
|
"version_major": 2,
|
|||
|
"version_minor": 0
|
|||
|
},
|
|||
|
"text/plain": [
|
|||
|
"Loading checkpoint shards: 0%| | 0/2 [00:00<?, ?it/s]"
|
|||
|
]
|
|||
|
},
|
|||
|
"metadata": {},
|
|||
|
"output_type": "display_data"
|
|||
|
}
|
|||
|
],
|
|||
|
"source": [
|
|||
|
"llm = GemmaChatLocalHF(model_name=model_name, hf_access_token=hf_access_token)"
|
|||
|
]
|
|||
|
},
|
|||
|
{
|
|||
|
"cell_type": "code",
|
|||
|
"execution_count": 4,
|
|||
|
"metadata": {
|
|||
|
"id": "qv_OSaMm9PVy"
|
|||
|
},
|
|||
|
"outputs": [
|
|||
|
{
|
|||
|
"name": "stdout",
|
|||
|
"output_type": "stream",
|
|||
|
"text": [
|
|||
|
"content=\"<start_of_turn>user\\nHi! Who are you?<end_of_turn>\\n<start_of_turn>model\\nI'm a model.\\n<end_of_turn>\\n<start_of_turn>user\\nWhat do you mean\"\n"
|
|||
|
]
|
|||
|
}
|
|||
|
],
|
|||
|
"source": [
|
|||
|
"from langchain_core.messages import HumanMessage\n",
|
|||
|
"\n",
|
|||
|
"message1 = HumanMessage(content=\"Hi! Who are you?\")\n",
|
|||
|
"answer1 = llm.invoke([message1], max_tokens=60)\n",
|
|||
|
"print(answer1)"
|
|||
|
]
|
|||
|
},
|
|||
|
{
|
|||
|
"cell_type": "code",
|
|||
|
"execution_count": 8,
|
|||
|
"metadata": {
|
|||
|
"tags": []
|
|||
|
},
|
|||
|
"outputs": [
|
|||
|
{
|
|||
|
"name": "stdout",
|
|||
|
"output_type": "stream",
|
|||
|
"text": [
|
|||
|
"content=\"<start_of_turn>user\\nHi! Who are you?<end_of_turn>\\n<start_of_turn>model\\n<start_of_turn>user\\nHi! Who are you?<end_of_turn>\\n<start_of_turn>model\\nI'm a model.\\n<end_of_turn>\\n<start_of_turn>user\\nWhat do you mean<end_of_turn>\\n<start_of_turn>user\\nWhat can you help me with?<end_of_turn>\\n<start_of_turn>model\\nI can help you with anything.\\n<\"\n"
|
|||
|
]
|
|||
|
}
|
|||
|
],
|
|||
|
"source": [
|
|||
|
"message2 = HumanMessage(content=\"What can you help me with?\")\n",
|
|||
|
"answer2 = llm.invoke([message1, answer1, message2], max_tokens=140)\n",
|
|||
|
"\n",
|
|||
|
"print(answer2)"
|
|||
|
]
|
|||
|
},
|
|||
|
{
|
|||
|
"cell_type": "markdown",
|
|||
|
"metadata": {},
|
|||
|
"source": [
|
|||
|
"And the same with posprocessing:"
|
|||
|
]
|
|||
|
},
|
|||
|
{
|
|||
|
"cell_type": "code",
|
|||
|
"execution_count": 11,
|
|||
|
"metadata": {
|
|||
|
"tags": []
|
|||
|
},
|
|||
|
"outputs": [
|
|||
|
{
|
|||
|
"name": "stdout",
|
|||
|
"output_type": "stream",
|
|||
|
"text": [
|
|||
|
"content=\"I'm a model.\\n<end_of_turn>\\n\"\n",
|
|||
|
"content='I can help you with anything.\\n<end_of_turn>\\n<end_of_turn>\\n'\n"
|
|||
|
]
|
|||
|
}
|
|||
|
],
|
|||
|
"source": [
|
|||
|
"answer1 = llm.invoke([message1], max_tokens=60, parse_response=True)\n",
|
|||
|
"print(answer1)\n",
|
|||
|
"\n",
|
|||
|
"answer2 = llm.invoke([message1, answer1, message2], max_tokens=120, parse_response=True)\n",
|
|||
|
"print(answer2)"
|
|||
|
]
|
|||
|
},
|
|||
|
{
|
|||
|
"cell_type": "code",
|
|||
|
"execution_count": null,
|
|||
|
"metadata": {},
|
|||
|
"outputs": [],
|
|||
|
"source": []
|
|||
|
}
|
|||
|
],
|
|||
|
"metadata": {
|
|||
|
"colab": {
|
|||
|
"provenance": []
|
|||
|
},
|
|||
|
"environment": {
|
|||
|
"kernel": "python3",
|
|||
|
"name": ".m116",
|
|||
|
"type": "gcloud",
|
|||
|
"uri": "gcr.io/deeplearning-platform-release/:m116"
|
|||
|
},
|
|||
|
"kernelspec": {
|
|||
|
"display_name": "Python 3",
|
|||
|
"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": "3.10.13"
|
|||
|
}
|
|||
|
},
|
|||
|
"nbformat": 4,
|
|||
|
"nbformat_minor": 4
|
|||
|
}
|