Skip to content

di ci: glm-5-2 1p 1d && 2p1d #1528

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
JiaoliangYu wants to merge 3 commits into
base: main
Choose a base branch
from
+72 −0
Open

di ci: glm-5-2 1p 1d && 2p1d #1528

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
02102aa
di ci: glm-5-2 1p 1d
Jul 9, 2026
31ee712
ci: add smoke suite for glm-5-2-fp8
Jul 9, 2026
5846064
ci: add 2p1d glm-5-2 case
Jul 9, 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
65 changes: 65 additions & 0 deletions .github/benchmark/models_atomesh.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -107,6 +107,71 @@ models:
fewshot: 3
limit: null

GLM-5.2-FP8:
backend: atom
model_path: "${ATOMESH_MODEL_ROOT}/GLM-5.2-FP8/"
precision: FP8
env:
common: {}
prefill: {}
decode: {}
server:
common_args:
trust_remote_code: true
kv_cache_dtype: fp8
block_size: 16
gpu_memory_utilization: 0.8
max_model_len: 32768
max_num_seqs: 256
max_num_batched_tokens: 32768
extra_args: ""
suites:
nightly:
- name: glm-5-2-fp8-1p1d-tp8
topology: 1p1d
pd_worker_layout: multi_node
nodes: "${ATOMESH_1P1D_NODES}"
isl: [8192]
osl: 1024
concurrency: [128, 256]
prefill:
workers: 1
tp: 8
decode:
workers: 1
tp: 8
benchmark:
wait_server_timeout: 4200
runner:
time_limit: "08:00:00"
run_eval: true
eval_concurrency: [64, 256]

- name: glm-5-2-fp8-2p1d-tp8
topology: 2p1d
pd_worker_layout: multi_node
nodes: "${ATOMESH_2P1D_NODES}"
isl: [8192]
osl: 1024
concurrency: [128, 256]
prefill:
workers: 2
tp: 8
decode:
workers: 1
tp: 8
benchmark:
wait_server_timeout: 4200
runner:
time_limit: "08:00:00"
run_eval: true
eval_concurrency: [64, 256]

accuracy:
task: gsm8k
fewshot: 3
limit: null

Kimi-K2.5-MXFP4:
backend: atom
model_path: "${ATOMESH_MODEL_ROOT}/Kimi-K2.5-MXFP4/"
Expand Down
7 changes: 7 additions & 0 deletions .github/workflows/atomesh-benchmark.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -42,6 +42,11 @@ on:
required: false
default: false
type: boolean
run_glm_5_2_fp8:
description: 'Run GLM-5.2-FP8'
required: false
default: false
type: boolean
run_kimi_k25_mxfp4:
description: 'Run Kimi-K2.5-MXFP4'
required: false
Expand Down Expand Up @@ -152,6 +157,7 @@ jobs:
SUITE: ${{ (github.event_name == 'schedule' || github.event_name == 'push') && 'nightly' || (github.event_name == 'workflow_dispatch' && inputs.suite || 'smoke') }}
RUN_ALL_MODELS: ${{ github.event_name == 'workflow_dispatch' && inputs.run_all_models || github.event_name != 'workflow_dispatch' }}
RUN_DEEPSEEK_V4_PRO: ${{ github.event_name == 'workflow_dispatch' && inputs.run_deepseek_v4_pro || false }}
RUN_GLM_5_2_FP8: ${{ github.event_name == 'workflow_dispatch' && inputs.run_glm_5_2_fp8 || false }}
RUN_KIMI_K25_MXFP4: ${{ github.event_name == 'workflow_dispatch' && inputs.run_kimi_k25_mxfp4 || false }}
RUN_MINIMAX_M3_MXFP4: ${{ github.event_name == 'workflow_dispatch' && inputs.run_minimax_m3_mxfp4 || false }}
RUN_MINIMAX_M3_MXFP8: ${{ github.event_name == 'workflow_dispatch' && inputs.run_minimax_m3_mxfp8 || false }}
Expand All @@ -164,6 +170,7 @@ jobs:
if [[ "${GITHUB_EVENT_NAME}" == "workflow_dispatch" && "${RUN_ALL_MODELS}" != "true" ]]; then
selected_models=()
[[ "${RUN_DEEPSEEK_V4_PRO}" == "true" ]] && selected_models+=("DeepSeek-V4-Pro")
[[ "${RUN_GLM_5_2_FP8}" == "true" ]] && selected_models+=("GLM-5.2-FP8")
[[ "${RUN_KIMI_K25_MXFP4}" == "true" ]] && selected_models+=("Kimi-K2.5-MXFP4")
[[ "${RUN_MINIMAX_M3_MXFP4}" == "true" ]] && selected_models+=("MiniMax-M3-MXFP4")
[[ "${RUN_MINIMAX_M3_MXFP8}" == "true" ]] && selected_models+=("MiniMax-M3-MXFP8")
Expand Down
Loading