# liguodongiot/llm-action architecture

> Architecture diagram of the GitHub repository liguodongiot/llm-action, generated by GitDiagram from its file tree, README and a sample of source files (last updated 2026-11-09). It is an AI-made overview: check details against the source.

- Interactive diagram: https://gitdiagram.com/liguodongiot/llm-action
- Repository: https://github.com/liguodongiot/llm-action

## Overview

This repository is an educational collection of LLM and AI engineering examples, not one integrated runtime. The README organizes tutorials around training, inference, compression, evaluation, data engineering, applications, algorithms, and domestic-hardware adaptation; the tree contains corresponding scripts. The principal user workflow is to choose a tutorial area and run its example, with training, serving, and performance measurement as distinct domain stages. The graph groups these capabilities rather than implying a shared application pipeline. Connections are limited to README-documented areas and excerpted behavior; many indexed scripts were not sampled.

Main entry points: [`llm-inference`](https://github.com/liguodongiot/llm-action/tree/main/llm-inference) (LLM Serving), [`llm-eval/llm-performance`](https://github.com/liguodongiot/llm-action/tree/main/llm-eval/llm-performance) (Inference Benchmarks), [`llm-inference/ascend/mindformers/mindsporelite-inference.py`](https://github.com/liguodongiot/llm-action/blob/main/llm-inference/ascend/mindformers/mindsporelite-inference.py) (MindSpore Lite), [`llm-inference/faster-transformer/bloom/firefly_lambada_1w_stat_token.py`](https://github.com/liguodongiot/llm-action/blob/main/llm-inference/faster-transformer/bloom/firefly_lambada_1w_stat_token.py) (LAMBADA Evaluation)

## Diagram (Mermaid)

```mermaid
flowchart TD

subgraph group_train["Training"]
  node_training["LLM Training"]
  node_peft["Efficient Tuning"]
  node_distributed["Distributed Training"]
  node_frameworks["AI Framework Demos"]
end

subgraph group_infer["Inference"]
  node_serving["LLM Serving"]
  node_mindspore["MindSpore Lite"]
  node_webapi["Web API"]
  node_vllmclient["vLLM Client<br/>[api_client.py]"]
  node_tritonclient["Triton Client<br/>[client.py]"]
end

subgraph group_compress["Compression"]
  node_qat["QAT Workflows"]
end

subgraph group_evaluate["Evaluation"]
  node_performance["Inference Benchmarks"]
  node_gpu_stats["GPU Monitoring<br/>[stat_gpu_memory.py]"]
  node_profiler["Profiling Tools<br/>[base-profiler.py]"]
  node_lambada["LAMBADA Evaluation"]
end

subgraph group_ecosystem["Data and Ecosystem"]
  node_dataset["SFT Templates"]
  node_algorithm["Model Algorithms"]
  node_application["LLM Applications<br/>[serve.py]"]
  node_localization["Hardware Adaptation"]
end

node_engineer(("Engineer"))

node_engineer -->|"explores tutorials"| node_training
node_engineer -->|"explores tutorials"| node_serving
node_engineer -->|"explores tutorials"| node_qat
node_engineer -->|"explores tutorials"| node_performance
node_engineer -->|"explores tutorials"| node_dataset
node_engineer -->|"explores tutorials"| node_application
node_engineer -->|"explores tutorials"| node_algorithm
node_engineer -->|"explores tutorials"| node_localization
node_engineer -->|"explores tutorials"| node_peft
node_engineer -->|"explores tutorials"| node_distributed
node_engineer -->|"explores tutorials"| node_frameworks
node_serving -->|"includes example"| node_mindspore
node_serving -->|"includes client"| node_vllmclient
node_serving -.->|"includes client"| node_tritonclient
node_mindspore -->|"serves through"| node_webapi
node_performance -->|"includes monitoring"| node_gpu_stats
node_performance -.->|"includes profiling"| node_profiler
node_lambada -.->|"measures inference"| node_performance

click node_training "https://github.com/liguodongiot/llm-action/tree/main/llm-train"
click node_peft "https://github.com/liguodongiot/llm-action/tree/main/llm-train/peft"
click node_distributed "https://github.com/liguodongiot/llm-action/tree/main/llm-train/pytorch/distribution"
click node_frameworks "https://github.com/liguodongiot/llm-action/tree/main/ai-framework"
click node_dataset "https://github.com/liguodongiot/llm-action/tree/main/llm-data-engineering/sft-dataset"
click node_serving "https://github.com/liguodongiot/llm-action/tree/main/llm-inference"
click node_mindspore "https://github.com/liguodongiot/llm-action/blob/main/llm-inference/ascend/mindformers/mindsporelite-inference.py"
click node_webapi "https://github.com/liguodongiot/llm-action/blob/main/llm-inference/web/fastapi/llm-qwen-mindspore-lite.py"
click node_vllmclient "https://github.com/liguodongiot/llm-action/blob/main/llm-inference/vllm/api_client.py"
click node_tritonclient "https://github.com/liguodongiot/llm-action/blob/main/llm-inference/triton/resnet50/client.py"
click node_algorithm "https://github.com/liguodongiot/llm-action/tree/main/llm-algo"
click node_application "https://github.com/liguodongiot/llm-action/blob/main/llm-application/langchain/serve.py"
click node_qat "https://github.com/liguodongiot/llm-action/tree/main/llm-compression/quantization/llm-qat"
click node_localization "https://github.com/liguodongiot/llm-action/tree/main/llm-localization/ascend"
click node_performance "https://github.com/liguodongiot/llm-action/tree/main/llm-eval/llm-performance"
click node_gpu_stats "https://github.com/liguodongiot/llm-action/blob/main/llm-eval/llm-performance/stat_gpu_memory.py"
click node_profiler "https://github.com/liguodongiot/llm-action/blob/main/llm-tools/base-profiler.py"
click node_lambada "https://github.com/liguodongiot/llm-action/blob/main/llm-inference/faster-transformer/bloom/firefly_lambada_1w_stat_token.py"

classDef toneNeutral fill:#f8fafc,stroke:#334155,stroke-width:1.5px,color:#0f172a
classDef toneBlue fill:#dbeafe,stroke:#2563eb,stroke-width:1.5px,color:#172554
classDef toneAmber fill:#fef3c7,stroke:#d97706,stroke-width:1.5px,color:#78350f
classDef toneMint fill:#dcfce7,stroke:#16a34a,stroke-width:1.5px,color:#14532d
classDef toneRose fill:#ffe4e6,stroke:#e11d48,stroke-width:1.5px,color:#881337
classDef toneIndigo fill:#e0e7ff,stroke:#4f46e5,stroke-width:1.5px,color:#312e81
classDef toneTeal fill:#ccfbf1,stroke:#0f766e,stroke-width:1.5px,color:#134e4a
class node_training,node_peft,node_distributed,node_frameworks toneBlue
class node_serving,node_mindspore,node_webapi,node_vllmclient,node_tritonclient toneAmber
class node_qat toneMint
class node_performance,node_gpu_stats,node_profiler,node_lambada toneRose
class node_dataset,node_algorithm,node_application,node_localization,node_engineer toneIndigo
```

## Components

### Training

- [LLM Training](https://github.com/liguodongiot/llm-action/tree/main/llm-train): `llm-train`
- [Efficient Tuning](https://github.com/liguodongiot/llm-action/tree/main/llm-train/peft): `llm-train/peft`
- [Distributed Training](https://github.com/liguodongiot/llm-action/tree/main/llm-train/pytorch/distribution): `llm-train/pytorch/distribution`
- [AI Framework Demos](https://github.com/liguodongiot/llm-action/tree/main/ai-framework): `ai-framework`

### Inference

- [LLM Serving](https://github.com/liguodongiot/llm-action/tree/main/llm-inference): `llm-inference`
- [MindSpore Lite](https://github.com/liguodongiot/llm-action/blob/main/llm-inference/ascend/mindformers/mindsporelite-inference.py): `llm-inference/ascend/mindformers/mindsporelite-inference.py`
- [Web API](https://github.com/liguodongiot/llm-action/blob/main/llm-inference/web/fastapi/llm-qwen-mindspore-lite.py): `llm-inference/web/fastapi/llm-qwen-mindspore-lite.py`
- [vLLM Client](https://github.com/liguodongiot/llm-action/blob/main/llm-inference/vllm/api_client.py): `llm-inference/vllm/api_client.py`
- [Triton Client](https://github.com/liguodongiot/llm-action/blob/main/llm-inference/triton/resnet50/client.py): `llm-inference/triton/resnet50/client.py`

### Compression

- [QAT Workflows](https://github.com/liguodongiot/llm-action/tree/main/llm-compression/quantization/llm-qat): `llm-compression/quantization/llm-qat`

### Evaluation

- [Inference Benchmarks](https://github.com/liguodongiot/llm-action/tree/main/llm-eval/llm-performance): `llm-eval/llm-performance`
- [GPU Monitoring](https://github.com/liguodongiot/llm-action/blob/main/llm-eval/llm-performance/stat_gpu_memory.py): `llm-eval/llm-performance/stat_gpu_memory.py`
- [Profiling Tools](https://github.com/liguodongiot/llm-action/blob/main/llm-tools/base-profiler.py): `llm-tools/base-profiler.py`
- [LAMBADA Evaluation](https://github.com/liguodongiot/llm-action/blob/main/llm-inference/faster-transformer/bloom/firefly_lambada_1w_stat_token.py): `llm-inference/faster-transformer/bloom/firefly_lambada_1w_stat_token.py`

### Data and Ecosystem

- [SFT Templates](https://github.com/liguodongiot/llm-action/tree/main/llm-data-engineering/sft-dataset): `llm-data-engineering/sft-dataset`
- [Model Algorithms](https://github.com/liguodongiot/llm-action/tree/main/llm-algo): `llm-algo`
- [LLM Applications](https://github.com/liguodongiot/llm-action/blob/main/llm-application/langchain/serve.py): `llm-application/langchain/serve.py`
- [Hardware Adaptation](https://github.com/liguodongiot/llm-action/tree/main/llm-localization/ascend): `llm-localization/ascend`

### Other components

- **Engineer**

## Connections

- Engineer → LLM Training: explores tutorials (evidence: `README.md`)
- Engineer → LLM Serving: explores tutorials (evidence: `README.md`)
- Engineer → QAT Workflows: explores tutorials (evidence: `README.md`)
- Engineer → Inference Benchmarks: explores tutorials (evidence: `README.md`)
- Engineer → SFT Templates: explores tutorials (evidence: `README.md`)
- Engineer → LLM Applications: explores tutorials (evidence: `README.md`)
- Engineer → Model Algorithms: explores tutorials (evidence: `README.md`)
- Engineer → Hardware Adaptation: explores tutorials (evidence: `README.md`)
- Engineer → Efficient Tuning: explores tutorials (evidence: `README.md`)
- Engineer → Distributed Training: explores tutorials (evidence: `README.md`)
- Engineer → AI Framework Demos: explores tutorials (evidence: `README.md`)
- LLM Serving → MindSpore Lite: includes example (evidence: `README.md`)
- LLM Serving → vLLM Client: includes client (evidence: `README.md`)
- LLM Serving → Triton Client: includes client
- MindSpore Lite → Web API: serves through (evidence: `llm-inference/web/fastapi/llm-qwen-mindspore-lite.py`)
- Inference Benchmarks → GPU Monitoring: includes monitoring (evidence: `README.md`)
- Inference Benchmarks → Profiling Tools: includes profiling
- LAMBADA Evaluation → Inference Benchmarks: measures inference
