Skip to content

AirLLM 2.8T LLM inference with single 4GB GPU

llama3.1 405b example

Gavin Li044146707a60
diff --git a/air_llm/examples/run_llama3.1_405B.ipynb b/air_llm/examples/run_llama3.1_405B.ipynb
new file mode 100644
index 0000000..f4d8c6b
--- /dev/null
+++ b/air_llm/examples/run_llama3.1_405B.ipynb
@@ -0,0 +1,68 @@
+{
+ "cells": [
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "id": "bfd29d17-9756-464f-b692-41ff20f41148",
+ "metadata": {},
+ "outputs": [],
+ "source": []
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "id": "1e07099a-03b9-49da-b0f8-c473b7f449eb",
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "from airllm import AutoModel\n",
+ "\n",
+ "MAX_LENGTH = 128\n",
+ "# could use hugging face model repo id:\n",
+ "model = AutoModel.from_pretrained(\"unsloth/Meta-Llama-3.1-405B-Instruct-bnb-4bit\")\n",
+ "\n",
+ "input_text = [\n",
+ " 'What is the capital of United States?',\n",
+ " ]\n",
+ "\n",
+ "input_tokens = model.tokenizer(input_text,\n",
+ " return_tensors=\"pt\", \n",
+ " return_attention_mask=False, \n",
+ " truncation=True, \n",
+ " max_length=MAX_LENGTH, \n",
+ " padding=False)\n",
+ " \n",
+ "generation_output = model.generate(\n",
+ " input_tokens['input_ids'].cuda(), \n",
+ " max_new_tokens=10,\n",
+ " use_cache=True,\n",
+ " return_dict_in_generate=True)\n",
+ "\n",
+ "output = model.tokenizer.decode(generation_output.sequences[0])\n",
+ "\n",
+ "print(output)\n"
+ ]
+ }
+ ],
+ "metadata": {
+ "kernelspec": {
+ "display_name": "Python 3 (ipykernel)",
+ "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": 5
+}