Skip to content

[Klaud Cold] Update DeepSeek-V4.1-Flash FP4 MI355X vLLM AgentX DSpark image to nightly-335dd77b / [Klaud Cold] 将 DeepSeek-V4.1-Flash FP4 MI355X vLLM AgentX DSpark 镜像更新至 nightly-335dd77b #4685

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
Klaud-Cold wants to merge 6 commits into
base: main
Choose a base branch
from
Open

[Klaud Cold] Update DeepSeek-V4.1-Flash FP4 MI355X vLLM AgentX DSpark image to nightly-335dd77b / [Klaud Cold] 将 DeepSeek-V4.1-Flash FP4 MI355X vLLM AgentX DSpark 镜像更新至 nightly-335dd77b #4685

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
6 commits
Select commit Hold shift + click to select a range
77a7868
chore(dsv41flash): bump MI355X vLLM AgentX DSpark image to nightly-33…
github-actions[bot] Oct 30, 2026
0bd3434
perf-changelog: dsv41flash MI355X vLLM AgentX DSpark image to nightly…
github-actions[bot] Oct 30, 2026
a84fe14
Merge main into klaud/auto-ce6a34d62edb49e0-15ee20e274843ef4 [skip-sw…
github-actions[bot] Oct 31, 2026
b949964
Merge main into klaud/auto-ce6a34d62edb49e0-15ee20e274843ef4 [skip-sw…
github-actions[bot] Oct 31, 2026
99e2b4e
Merge main into klaud/auto-ce6a34d62edb49e0-15ee20e274843ef4 [skip-sw…
github-actions[bot] Nov 1, 2026
e55d8a7
Merge main into klaud/auto-ce6a34d62edb49e0-15ee20e274843ef4 [skip-sw…
github-actions[bot] Nov 2, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion inferencex-e2e/configs/amd-master.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1202,7 +1202,7 @@ dsv41flash-fp4-mi355x-vllm-agentic-dspark:
# ROCm 10.0 nightly carrying vllm-project/vllm#59039 (decoder SWA bounded
# replay on ROCm), #61082 (mono decode on all 40 backbone layers) and #62131
# (mono decode on the DSpark draft layers), all switched by --monokernel.
image: vllm/vllm-openai-rocm:nightly-e68efc685ab1e6f1eda004ced7157e3ca411da2f@sha256:520e098f54ec96837228aaddea0a833cbf6374dbdffd480c1d65f26ffab7701b
image: vllm/vllm-openai-rocm:nightly-335dd77b09161c5d63139b4e7ae51bb9928305b1@sha256:3c634cc2c51eda7eb840a236b1f09a5ecf6a4b0cb92c4b8486e2158770507291
model: deepseek-ai/DeepSeek-V4.1-Flash
model-prefix: dsv41flash
runner: cluster:mi355x-amds
Expand Down
7 changes: 7 additions & 0 deletions inferencex-e2e/perf-changelog.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -11782,3 +11782,10 @@
- "C64 also sets PYTHONHASHSEED=42 and VLLM_ROCM_MINIMAX_M3_SP_PREFILL=1. Sequence-parallel prefill stays off on this checkpoint: at 8d0f05a0c it needs an MXFP8 kernel on layer.quant_method.kernel, which the Quark linear method does not expose, and every rank logs it as off."
- "Draft data types unchanged: the dspark/ draft loads with its own MXFP8 config (FP8 E4M3 attention and FFN projections with 1x32 UE8M0 scales; BF16 fc, Markov and confidence heads, norms) and an FP8 draft KV cache. Only its length changes, at C16/C32."
pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/4671

- config-keys:
- dsv41flash-fp4-mi355x-vllm-agentic-dspark
scenario-type: [agentic-coding]
description:
- "Update vLLM ROCm image from nightly-e68efc68 to nightly-335dd77b, pinned by digest."
pr-link: https://github.com/SemiAnalysisAI/InferenceX/pull/4685
Loading