Skip to content

[WIP] Kimi K2.6 agentic #2158

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
xinli-sw wants to merge 1 commit into
base: main
Choose a base branch
from
+4,675 −963
Open

[WIP] Kimi K2.6 agentic #2158

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
371 changes: 0 additions & 371 deletions .github/codeowner-signoff-verify-prompt.md
View file
Open in desktop

This file was deleted.

38 changes: 4 additions & 34 deletions .github/workflows/benchmark-multinode-tmpl.yml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -60,18 +60,6 @@ on:
prefill-tp:
required: true
type: string
prefill-pp:
required: false
type: string
default: '1'
prefill-dcp-size:
required: false
type: string
default: '1'
prefill-pcp-size:
required: false
type: string
default: '1'
prefill-ep:
required: true
type: string
Expand All @@ -89,18 +77,6 @@ on:
decode-tp:
required: true
type: string
decode-pp:
required: false
type: string
default: '1'
decode-dcp-size:
required: false
type: string
default: '1'
decode-pcp-size:
required: false
type: string
default: '1'
decode-ep:
required: true
type: string
Expand Down Expand Up @@ -197,17 +173,11 @@ env:

PREFILL_NUM_WORKERS: ${{ inputs.prefill-num-worker }}
PREFILL_TP: ${{ inputs.prefill-tp }}
PREFILL_PP_SIZE: ${{ inputs.prefill-pp }}
PREFILL_DCP_SIZE: ${{ inputs.prefill-dcp-size }}
PREFILL_PCP_SIZE: ${{ inputs.prefill-pcp-size }}
PREFILL_EP: ${{ inputs.prefill-ep }}
PREFILL_DP_ATTN: ${{ inputs.prefill-dp-attn }}

DECODE_NUM_WORKERS: ${{ inputs.decode-num-worker }}
DECODE_TP: ${{ inputs.decode-tp }}
DECODE_PP_SIZE: ${{ inputs.decode-pp }}
DECODE_DCP_SIZE: ${{ inputs.decode-dcp-size }}
DECODE_PCP_SIZE: ${{ inputs.decode-pcp-size }}
DECODE_EP: ${{ inputs.decode-ep }}
DECODE_DP_ATTN: ${{ inputs.decode-dp-attn }}

Expand All @@ -220,8 +190,8 @@ jobs:
timeout-minutes: 480
name: >-
${{ inputs.model-prefix }} ${{ inputs.precision }} ${{ inputs.runner }} ${{ inputs.framework == 'sglang' && 'sgl' || inputs.framework == 'dynamo-sglang' && 'dyn-sgl' || inputs.framework == 'sglang-disagg' && 'sgl-disagg' || inputs.framework }}
${{ inputs.prefill-num-worker }}P (TP${{ inputs.prefill-tp }}${{ inputs.prefill-pp != '' && inputs.prefill-pp != '1' && format('/PP{0}', inputs.prefill-pp) || '' }}${{ inputs.prefill-dcp-size != '' && inputs.prefill-dcp-size != '1' && format('/DCP{0}', inputs.prefill-dcp-size) || '' }}${{ inputs.prefill-pcp-size != '' && inputs.prefill-pcp-size != '1' && format('/PCP{0}', inputs.prefill-pcp-size) || '' }}${{ inputs.prefill-ep != '' && inputs.prefill-ep != '1' && format('/EP{0}', inputs.prefill-ep) || '' }}${{ inputs.prefill-dp-attn == 'true' && '/DPA' || '' }})
x ${{ inputs.decode-num-worker }}D (TP${{ inputs.decode-tp }}${{ inputs.decode-pp != '' && inputs.decode-pp != '1' && format('/PP{0}', inputs.decode-pp) || '' }}${{ inputs.decode-dcp-size != '' && inputs.decode-dcp-size != '1' && format('/DCP{0}', inputs.decode-dcp-size) || '' }}${{ inputs.decode-pcp-size != '' && inputs.decode-pcp-size != '1' && format('/PCP{0}', inputs.decode-pcp-size) || '' }}${{ inputs.decode-ep != '' && inputs.decode-ep != '1' && format('/EP{0}', inputs.decode-ep) || '' }}${{ inputs.decode-dp-attn == 'true' && '/DPA' || '' }})
${{ inputs.prefill-num-worker }}P (TP${{ inputs.prefill-tp }}${{ inputs.prefill-ep != '' && inputs.prefill-ep != '1' && format('/EP{0}', inputs.prefill-ep) || '' }}${{ inputs.prefill-dp-attn == 'true' && '/DPA' || '' }})
x ${{ inputs.decode-num-worker }}D (TP${{ inputs.decode-tp }}${{ inputs.decode-ep != '' && inputs.decode-ep != '1' && format('/EP{0}', inputs.decode-ep) || '' }}${{ inputs.decode-dp-attn == 'true' && '/DPA' || '' }})
${{ inputs.spec-decoding != 'none' && inputs.spec-decoding || '' }}
${{ inputs.kv-offloading != '' && inputs.kv-offloading != 'none' && format('{0} KV offload', inputs.kv-offloading) || '' }}
${{ inputs.kv-offload-backend != '' && inputs.kv-offload-backend != 'none' && inputs.kv-offload-backend != 'default' && inputs.kv-offload-backend || '' }}
Expand Down Expand Up @@ -251,8 +221,8 @@ jobs:
env:
RUNNER_NAME: ${{ runner.name }}
RUNNER_TYPE: ${{ inputs.runner }}
# Hash uniquely on all prefill/decode parallelism fields, worker counts, serving mode, concurrency, and runner.
RESULT_FILENAME: ${{ env.EXP_NAME }}_${{ env.PRECISION }}_${{ env.FRAMEWORK }}_prefill-tp${{ env.PREFILL_TP }}-pp${{ env.PREFILL_PP_SIZE }}-dcp${{ env.PREFILL_DCP_SIZE }}-pcp${{ env.PREFILL_PCP_SIZE }}-ep${{ env.PREFILL_EP }}-dp${{ env.PREFILL_DP_ATTN }}-nw${{ env.PREFILL_NUM_WORKERS }}_decode-tp${{ env.DECODE_TP }}-pp${{ env.DECODE_PP_SIZE }}-dcp${{ env.DECODE_DCP_SIZE }}-pcp${{ env.DECODE_PCP_SIZE }}-ep${{ env.DECODE_EP }}-dp${{ env.DECODE_DP_ATTN }}-nw${{ env.DECODE_NUM_WORKERS }}_disagg-${{ env.DISAGG }}_spec-${{ env.SPEC_DECODING }}_conc${{ join(fromJson(inputs.conc-list), 'x') }}_${{ runner.name }}
# Hash uniquely on {EXP_NAME}_{PRECISION}_{FRAMEWORK}_prefill-tp{}-ep{}-dp{}-nw{}_decode-tp{}-ep{}-dp{}-nw{}_disagg-{}_spec-{}_conc{}_{runner}
RESULT_FILENAME: ${{ env.EXP_NAME }}_${{ env.PRECISION }}_${{ env.FRAMEWORK }}_prefill-tp${{ env.PREFILL_TP }}-ep${{ env.PREFILL_EP }}-dp${{ env.PREFILL_DP_ATTN }}-nw${{ env.PREFILL_NUM_WORKERS }}_decode-tp${{ env.DECODE_TP }}-ep${{ env.DECODE_EP }}-dp${{ env.DECODE_DP_ATTN }}-nw${{ env.DECODE_NUM_WORKERS }}_disagg-${{ env.DISAGG }}_spec-${{ env.SPEC_DECODING }}_conc${{ join(fromJson(inputs.conc-list), 'x') }}_${{ runner.name }}
run: |
set -x
# Export RESULT_FILENAME early so it's available for artifact uploads even if cancelled
Expand Down
13 changes: 4 additions & 9 deletions .github/workflows/benchmark-tmpl.yml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -32,10 +32,6 @@ on:
tp:
required: true
type: string
pp:
required: false
type: string
default: '1'
dcp-size:
required: false
type: string
Expand Down Expand Up @@ -118,7 +114,6 @@ env:
FRAMEWORK: ${{ inputs.framework }}
PRECISION: ${{ inputs.precision }}
TP: ${{ inputs.tp }}
PP_SIZE: ${{ inputs.pp }}
DCP_SIZE: ${{ inputs.dcp-size }}
PCP_SIZE: ${{ inputs.pcp-size }}
EP_SIZE: ${{ inputs.ep }}
Expand Down Expand Up @@ -150,7 +145,7 @@ jobs:
timeout-minutes: 500
name: >-
${{ inputs.model-prefix }} ${{ inputs.precision }} ${{ inputs.runner }} ${{ inputs.framework == 'sglang' && 'sgl' || inputs.framework == 'dynamo-sglang' && 'dyn-sgl' || inputs.framework == 'sglang-disagg' && 'sgl-disagg' || inputs.framework }}
TP${{ inputs.tp }}${{ inputs.pp != '' && inputs.pp != '1' && format('/PP{0}', inputs.pp) || '' }}${{ inputs.dcp-size != '' && inputs.dcp-size != '1' && format('/DCP{0}', inputs.dcp-size) || '' }}${{ inputs.pcp-size != '' && inputs.pcp-size != '1' && format('/PCP{0}', inputs.pcp-size) || '' }}${{ inputs.ep != '' && inputs.ep != '1' && format('/EP{0}', inputs.ep) || '' }}${{ inputs.dp-attn && '/DPA' || '' }}
TP${{ inputs.tp }}${{ inputs.dcp-size != '' && inputs.dcp-size != '1' && format('/DCP{0}', inputs.dcp-size) || '' }}${{ inputs.pcp-size != '' && inputs.pcp-size != '1' && format('/PCP{0}', inputs.pcp-size) || '' }}${{ inputs.ep != '' && inputs.ep != '1' && format('/EP{0}', inputs.ep) || '' }}${{ inputs.dp-attn && '/DPA' || '' }}
${{ inputs.spec-decoding != 'none' && inputs.spec-decoding || '' }}
${{ inputs.kv-offloading != '' && inputs.kv-offloading != 'none' && format('{0} KV offload', inputs.kv-offloading) || '' }}
${{ inputs.kv-offload-backend != '' && inputs.kv-offload-backend != 'none' && inputs.kv-offload-backend != 'default' && inputs.kv-offload-backend || '' }}
Expand Down Expand Up @@ -191,13 +186,13 @@ jobs:
env:
RUNNER_NAME: ${{ runner.name }}
RUNNER_TYPE: ${{ inputs.runner }}
# Hash uniquely on {EXP_NAME}_{PRECISION}_{FRAMEWORK}_tp{}-pp{}-dcp{}-pcp{}-ep{}-dpa{}_disagg-{}_spec-{}_conc{}_{runner}
RESULT_FILENAME: ${{ env.EXP_NAME }}_${{ env.PRECISION }}_${{ env.FRAMEWORK }}_tp${{ env.TP }}-pp${{ env.PP_SIZE }}-dcp${{ env.DCP_SIZE }}-pcp${{ env.PCP_SIZE }}-ep${{ env.EP_SIZE }}-dpa${{ env.DP_ATTENTION }}_disagg-${{ env.DISAGG }}_spec-${{ env.SPEC_DECODING }}_conc${{ env.CONC }}_${{ runner.name }}
# Hash uniquely on {EXP_NAME}_{PRECISION}_{FRAMEWORK}_tp{}-dcp{}-pcp{}-ep{}-dpa{}_disagg-{}_spec-{}_conc{}_{runner}
RESULT_FILENAME: ${{ env.EXP_NAME }}_${{ env.PRECISION }}_${{ env.FRAMEWORK }}_tp${{ env.TP }}-dcp${{ env.DCP_SIZE }}-pcp${{ env.PCP_SIZE }}-ep${{ env.EP_SIZE }}-dpa${{ env.DP_ATTENTION }}_disagg-${{ env.DISAGG }}_spec-${{ env.SPEC_DECODING }}_conc${{ env.CONC }}_${{ runner.name }}
# Suppress per-job eval markdown from being appended to the step summary.
# We'll publish a single combined eval table in the collection job instead.
GITHUB_STEP_SUMMARY: ''
run: |
export GPU_COUNT=$((TP * PP_SIZE * PCP_SIZE))
export GPU_COUNT=$((TP * PCP_SIZE))
echo "GPU_COUNT=${GPU_COUNT}" >> "$GITHUB_ENV"

# Export RESULT_FILENAME early so it's available for artifact uploads even if cancelled
Expand Down
Loading
Loading