An educational implementation of an inference engine
Project description
mini-vllm
A minimal implementation of vLLM's core ideas: PagedAttention and continuous batching.
Benchmarks
Hardware: NVIDIA A100 (Modal)
Model: meta-llama/Llama-3.2-1B
Max tokens per request: 50
Prompt: "The meaning of life is"
mini-vllm Performance
| Batch Size | Duration | Total Tokens | Throughput |
|---|---|---|---|
| 1 | 4.59s | 50 | 10.90 tokens/sec |
| 4 | 1.01s | 250 | 248.48 tokens/sec |
| 16 | 1.20s | 1050 | 872.23 tokens/sec |
Comparison with vLLM
| Batch Size | mini-vllm | vLLM | Ratio (vLLM/mini) |
|---|---|---|---|
| 1 | 10.90 tokens/sec | 213.73 tokens/sec | 19.6x |
| 4 | 248.48 tokens/sec | 977.46 tokens/sec | 3.9x |
| 16 | 872.23 tokens/sec | 3510.41 tokens/sec | 4.0x |
References
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Filter files by name, interpreter, ABI, and platform.
If you're not sure about the file name format, learn more about wheel file names.
Copy a direct link to the current filters
File details
Details for the file mini_vllm-0.1.0.tar.gz.
File metadata
- Download URL: mini_vllm-0.1.0.tar.gz
- Upload date:
- Size: 109.0 kB
- Tags: Source
- Uploaded using Trusted Publishing? Yes
- Uploaded via: twine/6.1.0 CPython/3.13.7
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
1162c3b9a227ce948d9ec3a8151de14456cce8abf336779197b6e5784f228045
|
|
| MD5 |
7d7241961d458f9644e92ec551d9c1b6
|
|
| BLAKE2b-256 |
a2172709403b94adf569abca0b823d131892ac9d8533dbc975d6a26e0308c334
|
Provenance
The following attestation bundles were made for mini_vllm-0.1.0.tar.gz:
Publisher:
publish.yml on ubermenchh/mini-vllm
-
Statement:
-
Statement type:
https://in-toto.io/Statement/v1 -
Predicate type:
https://docs.pypi.org/attestations/publish/v1 -
Subject name:
mini_vllm-0.1.0.tar.gz -
Subject digest:
1162c3b9a227ce948d9ec3a8151de14456cce8abf336779197b6e5784f228045 - Sigstore transparency entry: 780860968
- Sigstore integration time:
-
Permalink:
ubermenchh/mini-vllm@ebf4380cf1a9b354188e22559033e4457e68534a -
Branch / Tag:
refs/tags/v0.1.0 - Owner: https://github.com/ubermenchh
-
Access:
public
-
Token Issuer:
https://token.actions.githubusercontent.com -
Runner Environment:
github-hosted -
Publication workflow:
publish.yml@ebf4380cf1a9b354188e22559033e4457e68534a -
Trigger Event:
release
-
Statement type:
File details
Details for the file mini_vllm-0.1.0-py3-none-any.whl.
File metadata
- Download URL: mini_vllm-0.1.0-py3-none-any.whl
- Upload date:
- Size: 11.3 kB
- Tags: Python 3
- Uploaded using Trusted Publishing? Yes
- Uploaded via: twine/6.1.0 CPython/3.13.7
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
1771964473a14d70737d5308aee4bc4bbc5c470edc35c578de52b76171079cec
|
|
| MD5 |
3a05e68567d1f2d3ebb4486376cab66c
|
|
| BLAKE2b-256 |
f9733ef988a129a5900c313cfd80c192d4e388172a6ebff3535ca08a3324b110
|
Provenance
The following attestation bundles were made for mini_vllm-0.1.0-py3-none-any.whl:
Publisher:
publish.yml on ubermenchh/mini-vllm
-
Statement:
-
Statement type:
https://in-toto.io/Statement/v1 -
Predicate type:
https://docs.pypi.org/attestations/publish/v1 -
Subject name:
mini_vllm-0.1.0-py3-none-any.whl -
Subject digest:
1771964473a14d70737d5308aee4bc4bbc5c470edc35c578de52b76171079cec - Sigstore transparency entry: 780860969
- Sigstore integration time:
-
Permalink:
ubermenchh/mini-vllm@ebf4380cf1a9b354188e22559033e4457e68534a -
Branch / Tag:
refs/tags/v0.1.0 - Owner: https://github.com/ubermenchh
-
Access:
public
-
Token Issuer:
https://token.actions.githubusercontent.com -
Runner Environment:
github-hosted -
Publication workflow:
publish.yml@ebf4380cf1a9b354188e22559033e4457e68534a -
Trigger Event:
release
-
Statement type:
