-
- Add a feature to reduce the amount of vertex and polygon data and generate a new FBX for Quests
-
Is it feasible to compress a real-time updated neural network and deploy it on terminal devices?
-
Hello!
I have been really excited about your work! I attempted to use Palu for model compression on the Qwen2 series models, but regardless of the compression rate I set, I seem to encounter signif…
-
Description:
In-memory compression for animation
Use case:
A lot, and long, animations take a lot of memory. Some basic runtime compression/key-reduction could help reduce that.
2017-06-26 14:0…
-
Dear Zhihao:
My test data set contains 300 images with a resolution of 3264*2488. The qarv_base model is used for encoding and decoding. The image size before compression is 637.24MB, and the image s…
-
Add a `doctr.models.utils` module to compress existing models and improve their latency / memory load for inference purposes on CPU. Some interesting leads to investigate:
- [x] FP conversion (#10)
…
-
**Describe the bug**
I tried to quantize Qwen1.5-MoE-A2.7B-Chat with w4a16 for vllm PR: https://github.com/vllm-project/vllm/pull/7766
raise error TypeError: forward() got multiple values for argume…
-
## 🐞Describing the bug
I have a pytorch model ,which I'm able to convert to CoreML and compress. If I add ClassificationConfig during conversion, I'm unable to compress the model. The error is:
`"id…
-
I was thinking, the hosted files (i.e. models) could use compression like brotli. Considering its all static files this could be done once instead on per request.
For example, [decoder_model_merge…
-
Previously we used boost::serialization + boost::iostreams for compressed portable binary archives. Now, with cereal portable binary archives, we need a lightweight alternative to boost::iostreams fo…