add one-key-start run.sh

README.md

@@ -38,16 +38,16 @@
 
 ```bash
 # 7B
-bash scripts/make-7b.sh
+bash run.sh 7b
 
-# 或 13B
-bash scripts/make-13b.sh
+# 13B
+bash run.sh 13b
 
-# 或 7B Chinese
-bash scripts/make-7b-cn.sh
+# 7B Chinese
+bash run.sh 7b-cn
 
-# 或 7B Chinese 4bit
-bash scripts/make-7b-cn-4bit.sh
+# 7B Chinese 4bit
+bash run.sh 7b-cn-4bit
 ```
 
 2. 选择适合你的命令，从 HuggingFace 下载 LLaMA2 或中文模型：
@@ -127,13 +127,13 @@ meta-llama
 
 ```bash
 # 7B
-bash scripts/run-7b.sh
-# 或 13B
-bash scripts/run-13b.sh
-# 或 Chinese 7B
-bash scripts/run-7b-cn.sh
-# 或 Chinese 7B 4BIT
-bash scripts/run-7b-cn-4bit.sh
+bash run.sh 7b
+# 13B
+bash run.sh 13b
+# Chinese 7B
+bash run.sh 7b-cn
+# Chinese 7B 4BIT
+bash run.sh 7b-cn-4bit
 ```
 
 模型运行之后，在浏览器中访问 `http://localhost7860` 或者 `http://你的IP地址:7860` 就可以开始玩了。

README_EN.md

@@ -39,16 +39,16 @@ Get started quickly, locally using the 7B or 13B models, using Docker.
 
 ```bash
 # 7B
-bash scripts/make-7b.sh
+bash run.sh 7b
 
-# OR 13B
-bash scripts/make-13b.sh
+# 13B
+bash run.sh 13b
 
-# OR 7B Chinese
-bash scripts/make-7b-cn.sh
+# 7B Chinese
+bash run.sh 7b-cn
 
-# OR 7B Chinese 4bit
-bash scripts/make-7b-cn-4bit.sh
+# 7B Chinese 4bit
+bash run.sh 7b-cn-4bit
 ```
 
 2. Download LLaMA2 Models from HuggingFace, or chinese models.
@@ -128,13 +128,13 @@ meta-llama
 
 ```bash
 # 7B
-bash scripts/run-7b.sh
-# OR 13B
-bash scripts/run-13b.sh
-# OR Chinese 7B
-bash scripts/run-7b-cn.sh
-# OR Chinese 7B 4BIT
-bash scripts/run-7b-cn-4bit.sh
+bash run.sh 7b
+# 13B
+bash run.sh 13b
+# Chinese 7B
+bash run.sh 7b-cn
+# Chinese 7B 4BIT
+bash run.sh 7b-cn-4bit
 ```
 
 enjoy, open `http://localhost7860` or `http://ip:7860` and play with the LLaMA2!

run.sh

@@ -0,0 +1,79 @@
+#!/bin/bash
+
+me_path="$(dirname "$(readlink -f "$0")")"
+
+case "$1" in
+7b)
+    image_name=soulteary/llama2
+    tag_name=7b
+    docker_file=docker/Dockerfile.7b
+    # MetaAI LLaMA2 Models (10~14GB vRAM)
+    mod_url=https://huggingface.co/meta-llama/Llama-2-7b-chat-hf
+    mod_path=meta-llama
+    ;;
+13b)
+    image_name=soulteary/llama2
+    tag_name=13b
+    docker_file=docker/Dockerfile.13b
+    # MetaAI LLaMA2 Models (10~14GB vRAM)
+    mod_url=https://huggingface.co/meta-llama/Llama-2-13b-chat-hf
+    mod_path=meta-llama
+    ;;
+7b-cn)
+    image_name=soulteary/llama2
+    tag_name=7b-cn
+    docker_file=docker/Dockerfile.7b-cn
+    # 或 Chinese LLaMA2 (10~14GB vRAM)
+    mod_url=https://huggingface.co/LinkSoul/Chinese-Llama-2-7b
+    mod_path=LinkSoul
+    ;;
+7b-cn-4bit)
+    image_name=soulteary/llama2
+    tag_name=7b-cn-4bit
+    docker_file=docker/Dockerfile.7b-cn-4bit
+    # 或 Chinese LLaMA2 4BIT (5GB vRAM)
+    mod_url=https://huggingface.co/soulteary/Chinese-Llama-2-7b-4bit
+    mod_path=soulteary
+    ;;
+*)
+    cat <<EOF
+Usage:
+
+    $0 7b
+    $0 7b-cn
+    $0 7b-cn-4bit
+    $0 13b
+
+EOF
+    exit 1
+    ;;
+esac
+## base image
+if docker image ls | grep -q "soulteary/llama2.*base"; then
+    echo "Found docker image soulteary/llama2:base, skip build."
+else
+    echo "Not found docker image soulteary/llama2:base, build..."
+    docker build -t soulteary/llama2:base "$me_path" -f docker/Dockerfile.base
+fi
+## run image
+if docker image ls | grep -q "${image_name}.*${tag_name}"; then
+    echo "Found docker image ${image_name}:${tag_name}, skip build."
+else
+    echo "Not found docker image ${image_name}:${tag_name}, build..."
+    docker build -t "${image_name}:${tag_name}" "$me_path" -f "$docker_file"
+fi
+## download model
+if [ -d "$me_path/$mod_path" ]; then
+    echo "Found model path: $me_path/$mod_path, skip download."
+else
+    echo "Not found model path: $me_path/$mod_path, download..."
+    mkdir -p "$me_path/$mod_path"
+    git lfs install
+    git clone "$mod_url" "$me_path/$mod_path/"
+fi
+
+## run
+docker run --gpus all --rm -it --ipc=host --ulimit memlock=-1 --ulimit stack=67108864 \
+    -p 7860:7860 \
+    -v "$me_path/$mod_path":"/app/$mod_path" \
+    "${image_name}:${tag_name}"