Comments (7)
I think you're doing things correct. Let me try to follow these steps in a fresh environment and see if I can reproduce.
from deepspeed-mii.
@aponte411 I'm not able to reproduce the protobuf error installing with a clean environment. pip
installs protobuf==3.20.3
for me. But regarding the other issue after you have set the environment variable... What version of transformers
do you have? I just checked and it looks like the latest transformers version is breaking our SD example. I would recommend version 4.24.0
until we can find a solution.
from deepspeed-mii.
@aponte411 try this PR on the DeepSpeed side, should fix the config problem you are seeing:
microsoft/DeepSpeed#2589
from deepspeed-mii.
@mrwyattii works when installing from master :)
from deepspeed-mii.
What's the version of protobuf you have in this environment? pip list | grep proto
Also, you can avoid this by setting the following environment variable (but as the error message notes, it could add latency to the message passing): export PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION=python
.
from deepspeed-mii.
@mrwyattii when I install using pip install deepspeed[sd] deepspeed-mii
it installs protobuf==4.21.11. Using this version, I ran into the TypeError: Descriptors cannot be created directly.
I tried downgrading but that may have messed it up. I tried again setting that env variable, but keeping this version 4.21.11, and that seems to have resolved the protobuf issue but now the AttributeError from this issue seems to pop back up.
I checked and I am using deepspeed==0.7.6 and deepspeed-mii==0.0.4. Am I missing something?
from deepspeed-mii.
Looks like the protobuf issue was due to my local setup. Creating a new venv with a fresh install fixes it! @mrwyattii I'm using transformers==4.24.0
but I am still seeing the DSUNet config issue.
from deepspeed-mii.
Related Issues (20)
- Cohere's Command-R model support HOT 1
- I can't tell from documentation if we're meant to use a chat template or if it's automatically implemented?
- Block when Call client inference in multiprocessing.Process HOT 3
- How can i use this library with langchain or llama_index? HOT 1
- inference_core_ops.so: undefined symbol: _Z19cuda_wf6af16_linearRN2at6TensorES1_S1_S1_S1_S1_iiii HOT 6
- Limit VRAM usage in serving the model HOT 2
- Any plans for produnction-ready services?
- Add support for DBRX
- [FEATURE REQUEST] Add Support for Qwen1.5-MoE Architecture in DeepSpeed-MII HOT 1
- how can I use deepspeed to split the model to submit GPU?
- Is openai compatible server still working?
- How do I launch the api on a graphics card other than cuda: 0 HOT 1
- How is the prompt segmentation specifically implemented for Dynamic SplitFuse? Is there any code implement or code snippet ?
- [FEATURE] Access to logits and final hidden layer HOT 1
- RuntimeError: The server socket has failed to listen on any local network address HOT 1
- Only running one replica even though setting many replicas
- [Problem]errno: 98 - Address already in use
- Performance with vllm
- error when using Qwen1.5-32B
- ValueError: Unsupported model type phi3
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 deepspeed-mii.