[BUG] HSTU backward does not support sm86 or sm89

[ShaobinChen-AH]

Describe the bug
hello, guys. I am running PYTHONPATH=${PYTHONPATH}:$(realpath ../) torchrun --nproc_per_node 1 --master_addr localhost --master_port 6000 ./training/pretrain_gr_ranking.py --gin-config-file ./training/configs/kuairand_1k_ranking.gin in my docker container. But it reports error "HSTU backward does not support sm86 or sm89" . I noticed that in another issue hstu currently does not support sm_89 and add support for it in recent request. My gpu is a6000 and its cuda capability is sm_86. can you help you to solve this problem?

Steps/Code to reproduce bug
Follow this guide to craft a minimal bug report. This helps us reproduce the issue you're having and resolve the issue more quickly.

Expected behavior
A clear and concise description of what you expected to happen.

Environment details (please complete the following information):

• Environment location: [Bare-metal, Docker, Cloud(specify cloud provider)]
• Method of recsys-examples install: [conda, Docker, pip, or from source]
  • If method of install is [Docker], provide docker pull & docker run commands used
• Run print_env.sh from the project root and paste the results here

Additional context
Add any other context about the problem here.

---

By submitting this issue, you agree to follow our code of conduct and our contributing guidelines.

[JacoCheung]

Hi @ShaobinChen-AH . We have already enabled sm86 in this PR ( in spite of the title ). Beware that the command to build hstu for sm86/sm89:

[ShaobinChen-AH]

@JacoCheung Hi! I changed the command you mentioned, and now it is "RUN cd /workspace/recsys-examples/corelib/hstu &&
HSTU_DISABLE_86OR89=FALSE HSTU_DISABLE_ARBITRARY=TRUE HSTU_DISABLE_LOCAL=TRUE HSTU_DISABLE_RAB=TRUE HSTU_DISABLE_DRAB=TRUE pip install . &&
cd hopper &&
HSTU_DISABLE_ARBITRARY=TRUE HSTU_DISABLE_SM8x=TRUE HSTU_DISABLE_LOCAL=TRUE HSTU_DISABLE_RAB=TRUE HSTU_DISABLE_DELTA_Q=FALSE HSTU_DISABLE_DRAB=TRUE pip install ."

Then, I recompile the program. However, it still reports error "RuntimeError: HSTU backward does not support sm86 or sm89."

Could you give me any suggestion?

[JacoCheung]

Hi @ShaobinChen-AH, which branch are you using? If your branch included my PR, the log message HSTU backward does not support sm86 or sm89 should be removed.

On the other hand, could you please check if you have installed and replaced the original hstu (if present) correctly?

[shijieliu]

close as no response.