Skip to content

[MOE][WIP] Integrate Qwen3.5 397B FP8 PTPC MOE Optimization for BS 1-32 #2949

Closed
sammysun0711 wants to merge 4 commits into
ROCm:Qwen3.5_devfrom
sammysun0711:qwen3.5_pyhip_moe_v3
Closed

[MOE][WIP] Integrate Qwen3.5 397B FP8 PTPC MOE Optimization for BS 1-32 #2949
sammysun0711 wants to merge 4 commits into
ROCm:Qwen3.5_devfrom
sammysun0711:qwen3.5_pyhip_moe_v3

Conversation

@sammysun0711
Copy link
Copy Markdown
Contributor

Motivation

Integrate Qwen3.5 397B FP8 PTPC MOE Optimization for BS 1-32

Technical Details

Test Plan

Change -t to specify num_token for [1, 2, 4, 8, 10, 12, 16, 32]
AITER_MOE_SMALL_BATCH=1 python3 op_tests/test_moe_2stage.py -q 2 -a silu -e 512 -k 10 -dim 4096,128 -p t --use-raw-for-ref t -t 1

Test Result

Submission Checklist

Signed-off-by: Xiake Sun <xiake.sun@amd.com>
Signed-off-by: Xiake Sun <xiake.sun@amd.com>
Signed-off-by: Xiake Sun <xiake.sun@amd.com>
@sammysun0711 sammysun0711 changed the title [MOE] Integrate Qwen3.5 397B FP8 PTPC MOE Optimization for BS 1-32 [MOE][WIP] Integrate Qwen3.5 397B FP8 PTPC MOE Optimization for BS 1-32 Apr 29, 2026
@sammysun0711 sammysun0711 marked this pull request as draft April 29, 2026 10:39
@sammysun0711
Copy link
Copy Markdown
Contributor Author

Follow up with #3103

@sammysun0711 sammysun0711 deleted the qwen3.5_pyhip_moe_v3 branch May 9, 2026 09:52
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant