gpu_poor - GitLite

KO | EN

gitlite — search

#python #ai #llm #go #android #machine-learning #docker #macos #awesome #cli #kubernetes #vue #ios #open-source

gpu_poor

(★ 1,404)

Calculate token/s & GPU memory requirement for any LLM. Supports llama.cpp/ggml/bnb/QLoRA quantization

파일 탐색기

최종 버전 다운로드 (.zip)

- all_configs.json
- asset-manifest.json
- favicon.ico
- index.html
- itsovermeme.png
- logo192.png
- logo512.png
- manifest.json
- robots.txt
- weback.jpg
- all_configs.json
- favicon.ico
- index.html
- itsovermeme.png
- logo192.png
- logo512.png
- manifest.json
- robots.txt
- weback.jpg
- App.css
- App.js
- App.test.js
- cpu_config.json
- gpu_config.json
- index.css
- index.js
- logo.svg
- reportWebVitals.js
- setupTests.js
- textBox.js
- textContent.js
.gitignore
package-lock.json
package.json
README.md
tailwind.config.js

# 설치 가이드

1. 코드 내려받기

git clone https://github.com/RahulSChand/gpu_poor

깃허브에서 프로젝트 코드 전체를 내 컴퓨터로 내려받습니다.

cd gpu_poor

방금 내려받은 프로젝트 폴더 안으로 이동합니다.

2. Node.js

쉬움 추천

사전 준비물

Git GitHub에서 프로젝트 코드를 내려받으려면 필요합니다.
Node.js npm 명령어를 쓰려면 Node.js가 설치되어 있어야 합니다. LTS(장기지원) 버전을 추천합니다.

npm install

package.json에 명시된 라이브러리들을 내려받아 설치합니다.

npm start

개발/실행 서버를 켭니다.

✅ 명령어 실행 후 터미널에 나타나는 주소(보통 http://localhost:3000 형태)를 브라우저에서 열어보세요.

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# Can my GPU run this LLM? & at what token/s?

![Made with](https://img.shields.io/badge/logo-javascript-blue?logo=javascript)

Calculates how much **GPU memory you need** and how much **token/s you can get** for any LLM & GPU/CPU.

Also breakdown of where it goes for training/inference with quantization (GGML/bitsandbytes/QLoRA) & inference frameworks (vLLM/llama.cpp/HF) supported

Link: **https://rahulschand.github.io/gpu_poor/**

### Demo

![new_upload](https://github.com/RahulSChand/gpu_poor/assets/16897807/14250f55-e886-4cc6-9aeb-08532382860c)

---

## Use cases/Features

#### 1. Calculate vRAM memory requirement 💾

<img width="643" alt="image" src="https://github.com/RahulSChand/gpu_poor/assets/16897807/29577394-0efd-42fb-aaf4-282e9a45d5db">

---

#### 2. Calculate ~token/s you can get ⏱️

<img width="647" alt="image" src="https://github.com/RahulSChand/gpu_poor/assets/16897807/77627c9b-5fdd-44cf-8b7d-452ff0563a8a">

---

#### 3. Approximate time for finetuning (ms per iteration) ⌛️

<img width="764" alt="image" src="https://github.com/RahulSChand/gpu_poor/assets/16897807/e5fd08a1-abb9-4e00-ad45-ba9bb15ec546">

---

For memory, output is total vRAM & its breakdown. It looks like below

```     
{
  "Total": 4000,
  "KV Cache": 1000,
  "Model Size": 2000,
  "Activation Memory": 500,
  "Grad & Optimizer memory": 0,
  "cuda + other overhead":  500
}
```

For token/s, additional info looks like below

```     
{
  "Token per second": 50,
  "ms per token": 20,
  "Prompt process time (s)": 5 s,
  "memory or compute bound?": Memory,
}
```

For training, output is time for each forward pass (in ms)

```     
{
  "ms per iteration (forward + backward)": 100,
  "memory or compute bound?": Memory,
}
```

---

### Purpose

made this to check if you can run a particular LLM on your GPU. Useful to figure out the following

1. How much token/s can I get?
2. How much total time to finetune? 
3. What quantization will fit on my GPU?
4. Max context length & batch-size my GPU can handle?
5. Which finetuning? Full? LoRA? QLoRA?
6. What is consuming my GPU memory? What to change to fit the LLM on GPU?

---

## Additional info + FAQ

### Can't we just look at the model size & figure this out?

Finding which LLMs your GPU can handle isn't as easy as looking at the model size because during inference (KV cache) takes susbtantial amount of memory. For example, with sequence length 1000 on llama-2-7b it takes 1GB of extra memory (using hugginface LlamaForCausalLM, with exLlama & vLLM this is 500MB). And during training both KV cache & activations & quantization overhead take a lot of memory. For example, llama-7b with bnb int8 quant is of size ~7.5GB but it isn't possible to finetune it using LoRA on data with 1000 context length even with RTX 4090 24 GB. Which means an additional 16GB memory goes into quant overheads, activations & grad memory.

### How reliable are the numbers?
The results can vary depending on your model, input data, cuda version & what quant you are using & it is impossible to predict exact values. I have tried to take these into account & make sure the results are within 500MB. Below table I cross-check 3b,7b & 13b model memories given by the website vs. what what I get on my RTX 4090 & 2060 GPUs. All values are within 500MB.

<img width="604" alt="image" src="https://github.com/RahulSChand/gpu_poor/assets/16897807/3d49a422-f174-4537-b5fa-42adc4b15a89">

### How are the values calculated?

`Total memory = model size + kv-cache + activation memory + optimizer/grad memory + cuda etc. overhead`
1. **Model size** = this is your `.bin` file size (divide it by 2 if Q8 quant & by 4 if Q4 quant).
2. **KV-Cache** = Memory taken by KV (key-value) vectors. Size =  `(2 x sequence length x hidden size)` _per layer_. For huggingface this `(2 x 2 x sequence length x hidden size)` _per layer_. In training the whole sequence is processed at once (therefore KV cache memory = 0)
3. **Activation Memory** = In forward pass every operation's output has to be stored for doing `.backward()`. For example if you do `output = Q * input` where `Q = (dim, dim)` and `input = (batch, seq, dim)` then output of shape `(batch, seq, dim)` will need to be stored (in fp16). This consumes the most memory in LoRA/QLoRA. In LLMs there are many such intermediate steps (after Q,K,V and after attention, after norm, after FFN1, FFN2, FFN3, after skip layer ....) Around 15 intermediate representations are saved _per layer_. 
4. **Optimizer/Grad memory** = Memory taken by `.grad` tensors & tensors associated with the optimizer (`running avg` etc.)
5. **Cuda etc. overhead** = Around 500-1GB memory is taken by CUDA whenever cuda is loaded. Also there are additional overheads when you use any quantization (like bitsandbytes). There is not straightforward formula here (I assume 650 MB overhead in my calculations for cuda overhead)

### Why are the results wrong?
Sometimes the answers might be very wrong in which case please open an issue here & I will try to fix it.

---

### TODO
1. Add support for vLLM for token/s
2. ~Add QLora~ ✅
3. ~Add way to measure approximste tokens/s you can get for a particular GPU~ ✅
4. ~Improve logic to get hyper-params from size~ (since hidden layer/intermediate size/number of layers can vary for a particular size) ✅
5. Add AWQ