You are viewing a plain text version of this content. The canonical link for it is here.

Posted to commits@tvm.apache.org by GitBox <gi...@apache.org> on 2020/11/26 02:03:16 UTC

[GitHub] [tvm] ZihengJiang opened a new pull request #6980: HAGO: An Automated Hardware-Aware Quantization Framework

ZihengJiang opened a new pull request #6980:
URL: https://github.com/apache/tvm/pull/6980


   This PR implements the automated quantization framework described in the [RFC](https://discuss.tvm.apache.org/t/rfc-search-based-automated-quantization/5483). More detail can be found in the [paper draft](https://www.ziheng.org/files/hago.pdf) also. We observe that HAGO achieves speedups of 2.09x, 1.97x, and 2.48x on Intel Xeon Cascade Lake CPUs, NVIDIA Tesla T4 GPUs, ARM Cortex-A CPUs on Raspberry Pi4 relative to full precision respectively, while maintaining high post-training quantization accuracy in each case.
   
   Still WIP.
   
   ## Highlights
   - Hardware Awareness 
   - Model-Agnostic Graph Transformation
   - Search-Based Optimization
   
   ## Results
   
   ### Accuracy for the MXNet Models
   ![image](https://user-images.githubusercontent.com/17693755/100298796-df917480-2f46-11eb-9f0f-9211e8f201cb.png)
   
   ### Accuracy for the TensorFlow Models
   ![image](https://user-images.githubusercontent.com/17693755/100298862-0c458c00-2f47-11eb-92c7-1d671013604b.png)
   
   ### Accuracy for the PyTorch Models
   ![image](https://user-images.githubusercontent.com/17693755/100298835-fdf77000-2f46-11eb-829d-92efe96d7820.png)
   
   ### Performance on x86 CPU
   ![image](https://user-images.githubusercontent.com/17693755/100298887-1d8e9880-2f47-11eb-92eb-b4b9fcfbe323.png)
   
   ### Performance on NVIDIA GPU
   ![image](https://user-images.githubusercontent.com/17693755/100298903-27b09700-2f47-11eb-959e-c2a2a588ebfb.png)
   
   ### Performance on ARM CPU
   ![image](https://user-images.githubusercontent.com/17693755/100298927-36974980-2f47-11eb-8b2b-6adbffb4c08a.png)
   


----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

For queries about this service, please contact Infrastructure at:
users@infra.apache.org

[GitHub] [tvm] ZihengJiang commented on pull request #6980: HAGO: An Automated Hardware-Aware Quantization Framework

Posted by GitBox <gi...@apache.org>.

ZihengJiang commented on pull request #6980:
URL: https://github.com/apache/tvm/pull/6980#issuecomment-734040804


   @FrozenGene Sounds good. Thanks!


----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

For queries about this service, please contact Infrastructure at:
users@infra.apache.org

[GitHub] [tvm] FrozenGene commented on pull request #6980: HAGO: An Automated Hardware-Aware Quantization Framework

Posted by GitBox <gi...@apache.org>.

FrozenGene commented on pull request #6980:
URL: https://github.com/apache/tvm/pull/6980#issuecomment-734040341


   @ZihengJiang  As this is WIP, is it better to mark this pr as draft pr? when it is ready, we could start to review it.


----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

For queries about this service, please contact Infrastructure at:
users@infra.apache.org

[GitHub] [tvm] masahi closed pull request #6980: HAGO: An Automated Hardware-Aware Quantization Framework

Posted by GitBox <gi...@apache.org>.

masahi closed pull request #6980:
URL: https://github.com/apache/tvm/pull/6980


   


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: commits-unsubscribe@tvm.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org