Comments (1)
vLLM does not guarantee bitwise correctness with transformers
In general, when working with GPUs, fp16 inference has numerical precision limitations. So running with different batch sizes or different implementations of the model will have different results.
With fp32, you should see vert similar results between transformers and vllm
from vllm.
Related Issues (20)
- [Bug]: Cannot load fp8 model of internlm2-chat-7b offline HOT 2
- [Performance]: GPU utilization is low when running large batches on H100 HOT 1
- [Usage]: Can't utilize all VRAM for context
- [Bug]: Failed to import from vllm._C with ImportError("/lib64/libc.so.6: version `GLIBC_2.32' not found HOT 4
- [New Model]: Mistral-Nemo HOT 2
- [Bug]: Distributed Inference and Serving HOT 3
- [Feature]: Any thoughts about MI50 support ? HOT 3
- [Bug]: gptq model fails on pascal gpu with long prompt
- Upgrade to numpy >= 2.0.0 HOT 1
- [New Model]: Proposal for implementing Llava-Next-Video HOT 2
- [Bug]: INFO 07-19 10:17:50 async_llm_engine.py:167] Aborted request 30d35945-d526-40bc-90c6-40ad73e639b9. INFO 07-19 10:17:50 async_llm_engine.py:49] Engine is gracefully shutting down. HOT 2
- [Feature]: Support Lora Adapter generated from mistral-finetune
- [Usage]: ncclSystemError: System call (e.g. socket, malloc) or external library call failed or device error. HOT 3
- [Bug]: CUDA Error when print HOT 2
- Fp8 support for mi300x HOT 5
- [Bug]: SpecDecode AsyncMetricsCollector _last_metrics_collect_time is never reset
- [Feature]: LLM2Vec (Fine-Tuned Embeddings) Support
- [Bug]: Crash possible with Pipeline Parallel when aborting requests HOT 1
- [Bug]: Intel GPU Test failing in CI
- [Feature]: Implementation of Sliding Window Attention for Full Context Support with Gemma-2
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from vllm.