deepspeedai / deepspeedai/DeepSpeed
[REQUEST] Benchmark custom config
Nobody has claimed this yet.
- Dominant language
- Python
- Stars
- 43.1k
- Forks
- 5k
- Avg merge
- 4d 15h
- Merged PRs (30d)
- 112
Description
Is your feature request related to a problem? Please describe.
Is there a way to benchmark deepspeed inference based on different model configs (that is not in hf). Since its just benchmark, I want to just be able to use random weights.
Describe the solution you'd like
A tutorial/example code to define custom config and run benchmarks
Describe alternatives you've considered
I tried to use ds_transformer and pass DeepSpeedInfereceConfig to run some benchmarks, but proving a bit difficult than I thought and I am not sure if its the best way to estimate DeepSpeed inference.
Additional context
Trying to benchmark DeepSpeed inference on custom model sizes and perhaps extend to custom architecture.
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Start by examining ds_transformer and DeepSpeedInferenceConfig, which the issue identifies as the attempted benchmarking path. Create a tutorial or example showing custom model configurations, random weights, and benchmark execution; done means users can benchmark custom model sizes without relying on a model from Hugging Face.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python
- Domain
- machine-learning, performance
- Issue type
- Feature
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Stale
- Clarity
- Mostly clear
- Newbie friendliness
- 30/100