### Install ms-swift from GitHub

Source: https://github.com/ruc-datalab/deepanalyze/blob/main/deepanalyze/ms-swift/README.md

Install the ms-swift library directly from its GitHub repository. This is the initial setup step.

```shell
pip install git+https://github.com/modelscope/ms-swift.git
```

```shell
git clone https://github.com/modelscope/ms-swift.git
cd ms-swift
pip install -e .
```

--------------------------------

### Quick Start: Fine-tuning Qwen2.5-7B-Instruct

Source: https://github.com/ruc-datalab/deepanalyze/blob/main/deepanalyze/SkyRL/skyrl-train/docs/GetStarted/Quick-start.md

Example command for 10 minutes of self-cognition fine-tuning of Qwen2.5-7B-Instruct on a single 3090 GPU.

```shell
swift-train --model_dir "Qwen/Qwen2.5-7B-Instruct" --num_train_epochs 0.1 --max_steps 100 --output_dir "./qwen2.5-7b-instruct-ft" --model_name "qwen2.5-7b-instruct" --lora_enable true --lora_r 8 --lora_alpha 16 --lora_dropout 0.05 --dataset_type "self-cognition" --dataset_path "/mnt/datasets/self-cognition/qwen2.5-7b-instruct" --sft_packing true --batch_size 2 --gradient_accumulation_steps 1 --learning_rate 1e-4 --weight_decay 0.01 --warmup_ratio 0.03 --lr_scheduler_type "cosine" --logging_steps 10 --save_strategy "steps" --save_steps 50 --evaluation_strategy "steps" --eval_steps 50 --do_eval true --report_to "tensorboard" --bf16 true --gradient_checkpointing true --deepspeed "ds_config.json"
```

--------------------------------

### Install Dependencies with uv

Source: https://github.com/ruc-datalab/deepanalyze/blob/main/deepanalyze/SkyRL/skyagent/README.md

Set up a virtual environment and install project dependencies using uv.

```bash
uv venv
uv sync
```

--------------------------------

### Install ms-swift

Source: https://github.com/ruc-datalab/deepanalyze/blob/main/deepanalyze/SkyRL/skyrl-train/docs/Instruction/采样.md

Install the ms-swift library with LLM support. Alternatively, clone the repository and install from source.

```shell
pip install ms-swift[llm] -U
```

```shell
git clone https://github.com/modelscope/ms-swift.git
cd ms-swift
pip install -e '.[llm]'
```

--------------------------------

### SWIFT Mirror Image Examples

Source: https://github.com/ruc-datalab/deepanalyze/blob/main/deepanalyze/ms-swift/docs/source/GetStarted/SWIFT-installation.md

Examples of SWIFT mirror image URLs for different versions and regions. These are used for environment setup.

```text
# swift3.5.3
modelscope-registry.cn-hangzhou.cr.aliyuncs.com/modelscope-repo/modelscope:ubuntu22.04-cuda12.4.0-py310-torch2.6.0-vllm0.8.5.post1-modelscope1.27.1-swift3.5.3
modelscope-registry.cn-beijing.cr.aliyuncs.com/modelscope-repo/modelscope:ubuntu22.04-cuda12.4.0-py310-torch2.6.0-vllm0.8.5.post1-modelscope1.27.1-swift3.5.3
modelscope-registry.us-west-1.cr.aliyuncs.com/modelscope-repo/modelscope:ubuntu22.04-cuda12.4.0-py310-torch2.6.0-vllm0.8.5.post1-modelscope1.27.1-swift3.5.3

# swift3.4.1.post1
modelscope-registry.cn-hangzhou.cr.aliyuncs.com/modelscope-repo/modelscope:ubuntu22.04-cuda12.4.0-py311-torch2.6.0-vllm0.8.5.post1-modelscope1.26.0-swift3.4.1.post1
modelscope-registry.cn-beijing.cr.aliyuncs.com/modelscope-repo/modelscope:ubuntu22.04-cuda12.4.0-py311-torch2.6.0-vllm0.8.5.post1-modelscope1.26.0-swift3.4.1.post1
modelscope-registry.us-west-1.cr.aliyuncs.com/modelscope-repo/modelscope:ubuntu22.04-cuda12.4.0-py311-torch2.6.0-vllm0.8.5.post1-modelscope1.26.0-swift3.4.1.post1
```

```text
# swift3.3.0.post1
modelscope-registry.cn-hangzhou.cr.aliyuncs.com/modelscope-repo/modelscope:ubuntu22.04-cuda12.4.0-py311-torch2.6.0-vllm0.8.3-modelscope1.25.0-swift3.3.0.post1
modelscope-registry.us-west-1.cr.aliyuncs.com/modelscope-repo/modelscope:ubuntu22.04-cuda12.4.0-py311-torch2.6.0-vllm0.8.3-modelscope1.25.0-swift3.3.0.post1

# swift3.2.2
modelscope-registry.us-west-1.cr.aliyuncs.com/modelscope-repo/modelscope:ubuntu22.04-cuda12.4.0-py311-torch2.5.1-modelscope1.25.0-swift3.2.2
```

--------------------------------

### Install ms-swift from source

Source: https://github.com/ruc-datalab/deepanalyze/blob/main/deepanalyze/SkyRL/skyrl-train/docs/Instruction/Sample.md

Installs the ms-swift library from its source code repository. This involves cloning the repository and then performing a local installation.

```shell
git clone https://github.com/modelscope/ms-swift.git
cd ms-swift
pip install -e '.[llm]'
```

--------------------------------

### Install SWIFT with Evaluation Support

Source: https://github.com/ruc-datalab/deepanalyze/blob/main/deepanalyze/SkyRL/skyrl-train/docs/Instruction/Evaluation.md

Install the ms-swift package with the 'eval' extra to enable evaluation capabilities. Alternatively, install from source.

```shell
pip install ms-swift[eval] -U
```

```shell
git clone https://github.com/modelscope/ms-swift.git
cd ms-swift
pip install -e '.[eval]'
```

--------------------------------

### Start and Stop DeepAnalyze WebUI

Source: https://github.com/ruc-datalab/deepanalyze/blob/main/README.md

Commands to install dependencies, start, and stop the DeepAnalyze WebUI. Ensure you have cloned the repository and downloaded the DeepAnalyze-8B model.

```bash
cd demo/chat/frontend
npm install
cd ..
bash start.sh
```

```bash
bash stop.sh
```

--------------------------------

### Start and Stop DeepAnalyze WebUI v2

Source: https://github.com/ruc-datalab/deepanalyze/blob/main/README.md

Commands to install dependencies, configure environment variables, start, and stop the DeepAnalyze WebUI v2. This version includes support for HeyWhale API keys and Docker-based sandboxed code execution.

```bash
cd demo/chat_v2/frontend 
npm install
cd ..
cp .env.example .env 
bash start.sh
```

```bash
bash stop.sh
```

--------------------------------

### Self-Cognition Fine-tuning Example

Source: https://github.com/ruc-datalab/deepanalyze/blob/main/deepanalyze/ms-swift/docs/source_en/GetStarted/Quick-start.md

Example command for performing 10 minutes of self-cognition fine-tuning of Qwen2.5-7B-Instruct on a single 3090 GPU.

```shell
swift sft --model_id "Qwen/Qwen2.5-7B-Instruct" --dataset_type "self_cognition" --max_epochs 1 --output_dir "./qwen2.5-7b-instruct-sft" --model_name "qwen2.5-7b-instruct" --num_gpus 1 --max_steps 100 --save_steps 10 --logging_steps 1 --eval_steps 10 --batch_size 1 --gradient_accumulation_steps 1 --learning_rate 1e-5 --lr_scheduler_type "cosine" --warmup_ratio 0.03 --weight_decay 0.01 --lora_enable true --lora_r 8 --lora_alpha 16 --lora_dropout 0.05 --lora_target_modules "all" --gradient_checkpointing true --deepspeed "zero2" --fp16 true
```

--------------------------------

### Fine-tune Qwen2.5-7B-Instruct

Source: https://github.com/ruc-datalab/deepanalyze/blob/main/deepanalyze/ms-swift/docs/source/GetStarted/Quick-start.md

Example of fine-tuning Qwen2.5-7B-Instruct for 10 minutes on a single 3090 GPU using ms-swift.

```shell
swift-train --model_dir /path/to/Qwen2.5-7B-Instruct --output_dir /path/to/output --num_train_epochs 0.1 --max_steps 100 --per_device_train_batch_size 2 --gradient_accumulation_steps 2 --learning_rate 1e-5 --model_name Qwen2.5-7B-Instruct --dataset_dir /path/to/dataset --use_lora True --lora_rank 8 --lora_alpha 32 --lora_dropout 0.05
```

--------------------------------

### Setup Retriever Environment

Source: https://github.com/ruc-datalab/deepanalyze/blob/main/deepanalyze/SkyRL/skyrl-train/docs/recipes/searchr1.md

Create and activate a dedicated Conda environment for the retrieval server. This includes installing specific versions of Python, PyTorch, and other necessary libraries like faiss-gpu.

```bash
# Create and activate the retriever environment with Python 3.10
conda create -n retriever python=3.10 -y
conda activate retriever

# Install PyTorch (with GPU support) and related libraries
conda install numpy==1.26.4 # needed to stop incompatible version of numpy from being installed via pip
pip install torch==2.6.0 torchvision==0.21.0 torchaudio==2.6.0 --index-url https://download.pytorch.org/whl/cu124

# Install other Python packages
pip install transformers datasets pyserini huggingface_hub

# Install the GPU version of faiss
conda install faiss-gpu==1.8.0 -c pytorch -c nvidia -y

# Install the API service framework
pip install uvicorn fastapi
```

--------------------------------

### Build Documentation

Source: https://github.com/ruc-datalab/deepanalyze/blob/main/deepanalyze/ms-swift/docs/README.md

Run this command in the root directory to build the documentation.

```shell
# in root directory:
make docs
```

--------------------------------

### Start Mock vLLM Server

Source: https://github.com/ruc-datalab/deepanalyze/blob/main/demo/mock_vllm/README.md

Navigate to the project directory and run this command to start the mock server. This server simulates vLLM's streaming response functionality.

```bash
cd mock_vllm
python start_mock_vllmserver.py
```

--------------------------------

### Install and Install Pre-commit Hooks

Source: https://github.com/ruc-datalab/deepanalyze/blob/main/deepanalyze/SkyRL/skyrl-train/docs/getting-started/development.md

Install pre-commit for linting and formatting. This command installs the package and then sets up the git hooks.

```bash
uv pip install pre-commit
pre-commit install
```

--------------------------------

### Create Virtual Environment and Install Dependencies

Source: https://github.com/ruc-datalab/deepanalyze/blob/main/deepanalyze/SkyRL/skyrl-train/README.md

Create a virtual environment using uv and install project dependencies, including optional extras like vLLM. Activate the environment afterwards.

```bash
# creates a venv at .venv/
uv sync --extra vllm 
source .venv/bin/activate
```

--------------------------------

### Install MS-SWIFT

Source: https://github.com/ruc-datalab/deepanalyze/blob/main/deepanalyze/SkyRL/skyrl-train/docs/Instruction/Pre-training-and-Fine-tuning.md

Install the MS-SWIFT library. If using deepspeed zero2/zero3, install the deepspeed library as well.

```shell
pip install ms-swift -U
```

```shell
# If using deepspeed zero2/zero3
pip install deepspeed -U
```

--------------------------------

### Install SWIFT from Source

Source: https://github.com/ruc-datalab/deepanalyze/blob/main/deepanalyze/ms-swift/docs/source_en/GetStarted/SWIFT-installation.md

Clone the SWIFT repository and install it locally using pip. The '[all]' extra can be included for full capabilities.

```shell
git clone https://github.com/modelscope/ms-swift.git
cd ms-swift
pip install -e .

# Full capabilities
# pip install -e '.[all]'
```

--------------------------------

### Deploying a Fine-Tuned Model with Full Parameters

Source: https://github.com/ruc-datalab/deepanalyze/blob/main/deepanalyze/SkyRL/skyrl-train/docs/Instruction/Pre-training-and-Fine-tuning.md

Command to start a deployment server for a model trained with full parameters. Use `--model` to specify the checkpoint directory and `--adapters` should be set to `None`.

```shell
CUDA_VISIBLE_DEVICES=0 \
swift deploy \
    --adapters output/vx-xxx/checkpoint-xxx \
    --infer_backend pt \
    --temperature 0 \
    --max_new_tokens 2048 \
    --served_model_name '<model-name>'
```

--------------------------------

### Install Dependencies from Requirements File

Source: https://github.com/ruc-datalab/deepanalyze/blob/main/docs/DeepAnalyze_API_Key_Usage_Guide.md

Install Python dependencies listed in the requirements.txt file. This is an alternative to installing packages individually.

```bash
pip install -r requirements.txt
```

--------------------------------

### Activate Virtual Environment and Install Dependencies with uv

Source: https://github.com/ruc-datalab/deepanalyze/blob/main/deepanalyze/SkyRL/skyrl-train/docs/getting-started/installation.md

Activate the created virtual environment and install project dependencies, including optional vllm support, using uv sync.

```bash
source <path_to_venv>/bin/activate
uv sync --active --extra vllm
```

--------------------------------

### Install Dependencies with Pip

Source: https://github.com/ruc-datalab/deepanalyze/blob/main/docs/DeepAnalyze_API_Key_Usage_Guide.md

Install the required Python packages for the DeepAnalyze application. This command installs all listed dependencies directly.

```bash
pip install gradio openai fastapi uvicorn pandas requests python-multipart
```

--------------------------------

### Initialize Trainer and Start Training

Source: https://github.com/ruc-datalab/deepanalyze/blob/main/deepanalyze/ms-swift/examples/notebook/qwen2_5-self-cognition/self-cognition-sft.ipynb

Instantiate the Seq2SeqTrainer with the prepared model, arguments, and datasets, then begin the training process. This step also enables gradient checkpointing for memory efficiency.

```python
# Get the trainer and start the training.
model.enable_input_require_grads()  # Compatible with gradient checkpointing
trainer = Seq2SeqTrainer(
    model=model,
    args=training_args,
    data_collator=template.data_collator,
    train_dataset=train_dataset,
    eval_dataset=val_dataset,
    template=template,
)
trainer.train()

last_model_checkpoint = trainer.state.last_model_checkpoint
logger.info(f'last_model_checkpoint: {last_model_checkpoint}')
```

--------------------------------

### Start SWIFT Web UI

Source: https://github.com/ruc-datalab/deepanalyze/blob/main/deepanalyze/ms-swift/docs/source_en/GetStarted/Web-UI.md

Use this command to launch the SWIFT Web UI for training and inference. Specify the language with the --lang argument.

```shell
swift web-ui --lang zh
# or en
swift web-ui --lang en
```

--------------------------------

### Run Example Scripts

Source: https://github.com/ruc-datalab/deepanalyze/blob/main/API/README.md

Execute these example scripts to test the API functionality. One uses the requests library, and the other uses the OpenAI library.

```bash
cd example
python exampleRequest.py          #  requests example
python exampleOpenAI.py    # OpenAI library example
```

--------------------------------

### Install Megatron-SWIFT Dependencies

Source: https://github.com/ruc-datalab/deepanalyze/blob/main/deepanalyze/SkyRL/skyrl-train/docs/Instruction/Megatron-SWIFT-Training.md

Install necessary Python packages for Megatron-SWIFT training. Ensure you use the recommended PyTorch versions. Refer to the provided GitHub issues for potential installation troubleshooting.

```shell
pip install pybind11
# transformer_engine
# If an installation error occurs, you can refer to this issue for resolution: https://github.com/modelscope/ms-swift/issues/3793
pip install git+https://github.com/NVIDIA/TransformerEngine.git@release_v2.3

# apex
git clone https://github.com/NVIDIA/apex
cd apex
# https://github.com/modelscope/ms-swift/issues/4176
git checkout e13873debc4699d39c6861074b9a3b2a02327f92
pip install -v --disable-pip-version-check --no-cache-dir --no-build-isolation --config-settings "--build-option=--cpp_ext" --config-settings "--build-option=--cuda_ext" ./

# megatron-core
pip install git+https://github.com/NVIDIA/Megatron-LM.git@core_r0.12.0
```

--------------------------------

### Install SWIFT from Source Code

Source: https://github.com/ruc-datalab/deepanalyze/blob/main/deepanalyze/SkyRL/skyrl-train/docs/GetStarted/SWIFT-installation.md

Clone the SWIFT repository and install it in editable mode using pip. The '[all]' extra can be included for full capabilities.

```shell
git clone https://github.com/modelscope/ms-swift.git
cd ms-swift
pip install -e .
```

```shell
pip install -e '.[all]'
```

--------------------------------

### Multimodal Model Inference Example

Source: https://github.com/ruc-datalab/deepanalyze/blob/main/deepanalyze/ms-swift/docs/source/Instruction/Inference-and-deployment.md

Example script for running multimodal inference.

```shell
bash examples/app/mllm.sh
```

--------------------------------

### OpenAI Client Example

Source: https://github.com/ruc-datalab/deepanalyze/blob/main/deepanalyze/ms-swift/docs/source/Instruction/Inference-and-deployment.md

Example of using the OpenAI client for inference with ms-swift.

```python
python examples/deploy/client/llm/openai_client.py
```

--------------------------------

### Start Web-UI for Training and Deployment

Source: https://github.com/ruc-datalab/deepanalyze/blob/main/deepanalyze/ms-swift/README.md

Launch the Web-UI for a zero-threshold training and deployment interface. The UI language can be set using the SWIFT_UI_LANG environment variable.

```shell
SWIFT_UI_LANG=en swift web-ui
```

--------------------------------

### load function example

Source: https://github.com/ruc-datalab/deepanalyze/blob/main/deepanalyze/ms-swift/docs/README.md

Example of the `load` function with Google style docstring.

```APIDOC
    ```python
    def load(file, file_format=None, **kwargs):
        """Load data from json/yaml/pickle files.

        This method provides a unified api for loading data from serialized files.

        Args:
            file (str or :obj:`Path` or file-like object): Filename or a file-like
                object.
            file_format (str, optional): If not specified, the file format will be
                inferred from the file extension, otherwise use the specified one.
                Currently supported formats include "json", "yaml/yml".

        Examples:
            >>> load('/path/of/your/file')  # file is stored in disk
            >>> load('https://path/of/your/file')  # file is stored on internet
            >>> load('oss://path/of/your/file')  # file is stored in petrel

        Returns:
            The content from the file.
        """
    ```
```

--------------------------------

### Start SearchR1 Training Run

Source: https://github.com/ruc-datalab/deepanalyze/blob/main/deepanalyze/SkyRL/skyrl-train/docs/recipes/searchr1.md

From your base environment, set your Weights & Biases API key and execute the training script. This script uses 'uv' to package dependencies separately from the retriever environment.

```bash
export WANDB_API_KEY=your_wandb_api_key
bash examples/search/run_search.sh
```

--------------------------------

### Install SkyRL-Gym from PyPI

Source: https://github.com/ruc-datalab/deepanalyze/blob/main/deepanalyze/SkyRL/skyrl-gym/README.md

Install the latest release of the skyrl-gym library using pip.

```bash
pip install skyrl-gym
```

--------------------------------

### Basic Sampling with SWIFT

Source: https://github.com/ruc-datalab/deepanalyze/blob/main/deepanalyze/SkyRL/skyrl-train/docs/Instruction/Sample.md

Demonstrates the basic usage of the SWIFT sampling command-line tool. Specify the model, sampler engine, number of return sequences, and the dataset to generate sampled data.

```shell
swift sample --model LLM-Research/Meta-Llama-3.1-8B-Instruct --sampler_engine pt --num_return_sequences 5 --dataset AI-ModelScope/alpaca-gpt4-data-zh#5
```

--------------------------------

### Install Dependencies and Patch for mmengine Error

Source: https://github.com/ruc-datalab/deepanalyze/blob/main/deepanalyze/SkyRL/skyrl-train/docs/Instruction/Frequently-asked-questions.md

Addresses mmengine errors in Python 3.11 environments by suggesting a Python 3.10 environment or installing dependencies and a specific patch. Ensure all dependencies are installed before applying the patch.

```shell
pip3 install evalscope[all]
pip3 install https://modelscope-open.oss-cn-hangzhou.aliyuncs.com/package/evalscope-0.5.3.post1-py3-none-any.whl
```

--------------------------------

### InfoNCE Loss Configuration Examples

Source: https://github.com/ruc-datalab/deepanalyze/blob/main/deepanalyze/SkyRL/skyrl-train/docs/BestPractices/Embedding.md

Examples of JSON objects demonstrating different configurations for InfoNCE loss, including query, response, and image paths. These examples illustrate how to structure data for training embedding models.

```json
{"query": "<image>", "response": "sentence", "images": "/some/images.jpg"}
```

```json
{"query": "<image>sentence1", "response": "<image>sentence2", "rejected_response": ["<image>sentence1", "<image>sentence2"], "images": ["/some/images.jpg", "/some/images.jpg", "/some/images.jpg", "/some/images.jpg"]}
```