Skip to content
Closed
Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
165 commits
Select commit Hold shift + click to select a range
7eea23f
cp pr5373 pr5379 pr5410 (#5411)
bukejiyu Dec 5, 2025
707d1a1
[New][RL] Support Rollout Routing Replay (#5405) (#5408)
gongshaotian Dec 8, 2025
7926add
[Cherry-Pick][Loader][BugFix] Fix some parameters place on CPU in Pad…
SigureMo Dec 8, 2025
1dceb1c
Update setup.py
Jiang-Jia-Jun Dec 8, 2025
d4c16aa
[BugFix][Cherry-Pick] fix can not enter into cuda graph (#5423)
zhoutianzi666 Dec 8, 2025
31436a3
[Cherry-Pick] [BugFix] [RL] remove shutdown_process_group/restart_pro…
liyonghua0910 Dec 8, 2025
4b9e2c5
[BugFix] 0 not into cuda graph to save memory (#5426) (#5432)
zhoutianzi666 Dec 9, 2025
2c55bbc
support dynamic load for normal (#5437)
gzy19990617 Dec 9, 2025
b491dcd
[Optimization] compulte real max_logprobs in batch (#5430) (#5448)
ckl117 Dec 9, 2025
e9174f2
commit (#5452)
zhoutianzi666 Dec 9, 2025
1776d41
fix limit_thinking bug (#5469)
yuanlehome Dec 10, 2025
c5c43e3
fix attention bug in spec decoding (#5481)
freeliuzc Dec 10, 2025
bcde798
[CI][XPU] ep+prefix cache+chunk prefill (#5490)
zccjjj Dec 10, 2025
7019afb
[BugFix] fix instability after clearing weight (#5487)
liyonghua0910 Dec 11, 2025
b435639
[CI] disable test_cuda_graph_dynamic_subgraph.py in unit_test
EmmonsCurse Dec 11, 2025
71781b5
RL fix (#5505)
bukejiyu Dec 11, 2025
4e5e36e
[[Cherry-Pick][BugFix] fix hung when n>1 and --enable-logprob (#5492)…
ckl117 Dec 11, 2025
12e0206
[Cherry-Pick] [BugFix] [RL] skip model executing after clearing/updat…
liyonghua0910 Dec 12, 2025
5bdef76
[Feature][Optimization] Qwen Support Dynamic block_wise_fp8 cache (#5…
ckl117 Dec 15, 2025
0fa40f5
Fix bug for caching output when preempted (#5510)
rainyfly Dec 15, 2025
99b4024
[Cherry-Pick][BugFix] fix dynamic c8 in v1 loader(#5562) (#5519)
yuanlehome Dec 15, 2025
9f74233
【NewFeature】support load fp8 weight (#5566)
gzy19990617 Dec 16, 2025
53158b7
[Cherry-Pick][CI] Adape unit_test due to incompatibility change(#5578…
EmmonsCurse Dec 16, 2025
c19af49
[Cherry-Pick][RL] R3 Support RDMA Store(#5467) (#5468)
gongshaotian Dec 17, 2025
a7359d1
[Cherry-Pick][CI]Support different inferseed in speculate decoding(#5…
freeliuzc Dec 17, 2025
d67b64d
add detoken switch (#5463) (#5572)
qwes5s5 Dec 17, 2025
d7d633a
[Cherry-Pick][CI]Fix write qknorm cache bug in speculative decoding(#…
freeliuzc Dec 17, 2025
e56c4dd
[Cherry-Pick] Support for request-level speculative decoding metrics …
Deleter-D Dec 17, 2025
5300e73
[Others] Maintain the mtp branch temporarily. (#5446) (#5621)
carryyu Dec 17, 2025
a30a5b4
[Model] tp+ep support v1_loader (#5600)
Wanglongzhi2001 Dec 18, 2025
0cb9ad1
[Cherry-Pick][BugFix] fix speculate_limit_thinking_content_length #5…
yuanlehome Dec 18, 2025
646d1a0
[Cherry-Pick][RL]Support loading weights via the load_weights functio…
bukejiyu Dec 18, 2025
9c55bc3
[Cherry-Pick][BugFix] fix rl model_weights_signal to support tp>1 #56…
yuanlehome Dec 18, 2025
2aa88d3
[Cherry-Pick][RL]Fix RL load_weights #5642 (#5643)
bukejiyu Dec 19, 2025
23bfd28
[Cherry-Pick][BugFix] cp fix_cpu_cache_bugs(#5544) (#5577)
kevincheng2 Dec 19, 2025
b3f7881
update rl signal (#5650)
yuanlehome Dec 19, 2025
a9bb24b
[XPU]logprob bug (#5636)
qw86972190 Dec 19, 2025
e10c5d5
cp fix eb5 prefix cache bug (#5644)
kevincheng2 Dec 19, 2025
dd0014b
del core (#5659)
bukejiyu Dec 19, 2025
abf53b1
[BugFix] Fix custom_all_reduce overflow (#5662) (#5667)
ckl117 Dec 19, 2025
ea16c82
[Cherry-Pick] [RL] provide options for whether shutdown comm group af…
liyonghua0910 Dec 19, 2025
9006508
[BugFix] fix rl signal (#5678)
yuanlehome Dec 22, 2025
eb309e5
[XPU]Set top_p=0.0 by default on XPU to optimize performance (#5688)
ddchenhao66 Dec 23, 2025
ceafd75
[Speculative Decoding]Support multi-step mtp with cudagraph (#5624) (…
freeliuzc Dec 23, 2025
9ff99d2
[BugFix] fix double shutdown of comm group when rank0 clears weights …
liyonghua0910 Dec 23, 2025
f50988d
[Cherry-Pick][CI] Revert adapt vl_model baseline changes due to Paddl…
EmmonsCurse Dec 24, 2025
e51af01
[Cherry-Pick][Feature] Entropy calculation support #5692 (#5731)
Deleter-D Dec 24, 2025
e293c8c
check (#5736) (#5747)
ckl117 Dec 24, 2025
70163dd
[Cherry-Pick][CI] Refactor RL tests to reuse upload_clear(#5741) (#5755)
EmmonsCurse Dec 24, 2025
6945f87
[Bug fix] Set enable_cache_output as false by default (#5752)
rainyfly Dec 24, 2025
65e00c9
[Cherry-Pick][CI] Fix ci_image_update error of no depends
EmmonsCurse Dec 24, 2025
fc3bccc
[Cherry-Pick][Others]upgrade paddleformer to 0.4.0 #5599 (#5716)
bukejiyu Dec 24, 2025
7c62626
[Cherry-Pick][Loader]Fix bug in MTP weight loading #5744 (#5745)
bukejiyu Dec 25, 2025
d0c5bce
[cherry-pick] support FA3 in mixed mode and support Qwen3 rope (#5655)
zhoutianzi666 Dec 25, 2025
9a8e215
[BugFix][Cherry-Pick] cp fix logprob bug(#5604) (#5770)
kevincheng2 Dec 25, 2025
c170fc4
[FDConfig][Cherry-Pick] Cp disable mm chunked(#5774) (#5775)
kevincheng2 Dec 26, 2025
c78c3be
[BugFix] Fix preemption out of real_bsz (#5806)
rainyfly Dec 29, 2025
df775c2
[BugFix] Fix process_response_dict to support async in serving_comple…
kxz2002 Dec 29, 2025
a52c82a
[Model] support mode config for expert_dispatch (#5749)
Wanglongzhi2001 Dec 29, 2025
b2bd259
[Cherry-Pick][BugFix] Fix _disable_sequence_parallel_moe_if_needed#57…
yuanlehome Dec 29, 2025
aff3e67
support glm fa3 (#5586) (#5810)
ckl117 Dec 29, 2025
ca4ccf2
[BugFix] fix shm opened but not closed in set_data_ipc (#5827)
liyonghua0910 Dec 29, 2025
8345027
[RL] add lm_head_fp32 in RolloutModelConfig (#5824)
tianhaodongbd Dec 30, 2025
0d29f6d
[Cherry-Pick][BugFix] Fix entropy bugs (#5818) (#5819)
Deleter-D Dec 30, 2025
a247260
eb5 mm skip prefix cache (#5839)
kevincheng2 Dec 30, 2025
f33e642
[Cherry-Pick][Speculative Decoding] Optimize draft logprob (#5842) (#…
Deleter-D Dec 31, 2025
6380093
[Cherry-Pick] [BugFix] fix cache manager not launched in case of mtp …
liyonghua0910 Dec 31, 2025
20024b8
[Cherry-Pick][BugFix] cp skip_mm_revert(#5848) (#5849)
kevincheng2 Dec 31, 2025
9a7eb33
[Cherry-Pick][Optimization] Optimization for gather_logprob by 10GB (…
ckl117 Dec 31, 2025
3e04e43
[Cherry-Pick][XPU]MAX_BSZ aligns gpu settings and disable prefix cach…
ddchenhao66 Jan 4, 2026
180e6f9
[XPU][CI]Release ci update (#5687)
plusNew001 Jan 4, 2026
2a71e42
[Cherry-Pick][CI] Fix archive URL injection and add retry(#5725,#5828…
EmmonsCurse Jan 4, 2026
9de6ae3
[Cherry-Pick][APIServer][Feature] Add configurable worker health chec…
Copilot Jan 5, 2026
d624c52
[RL] Change 'model' to the instance variable 'tmp_model' (#5873)
tianhaodongbd Jan 5, 2026
0ee6391
support fa3 qwen-vl rope (#5869) (#5877)
ckl117 Jan 5, 2026
c9a806d
fix speculate metrics bug (#5875)
Deleter-D Jan 5, 2026
dcb0cce
[Speculative Decoding] Fix attn_mask_offset for multi-step MTP in mix…
freeliuzc Jan 5, 2026
0f008b8
[Cherry-Pick][OPs] ep_moe_expert_dispatch.cu dispatch num_experts_per…
yuanlehome Jan 6, 2026
f3ebd64
[Cherry-Pick] [KVCache] launch cache transfer processes only if hiera…
liyonghua0910 Jan 6, 2026
682e1ab
[Cherry-Pick] [BugFix] fix mtp cache attaching for pd disaggregation …
liyonghua0910 Jan 6, 2026
44e44ab
[Bugfix]fix model weight signal tensor num (#5899)
gzy19990617 Jan 6, 2026
3002334
[Cherry-Pick] [XPU]Cherry-pick Support ZMQ logprobs(#5628) (#5852)
iosmers Jan 7, 2026
ed3db9d
logging switch (#5765)
qwes5s5 Jan 7, 2026
939dfa4
[BugFix][Cherry-Pick] Cp fix eb5 prefix cache(#5879) (#5881)
kevincheng2 Jan 7, 2026
fb59f56
support multi-step draft-model with cudagraph (#5898)
freeliuzc Jan 7, 2026
7cdffce
[Cherry Pick][XPU][CI] Add logprobs Case (#5907)
plusNew001 Jan 7, 2026
0b630fc
[Cherry-Pick] [BugFix] fix mtp split kv attetion (#5921)
lizhenyun01 Jan 7, 2026
1e8de96
[Optim][Cherry-pick] Reduce preemption occurrence when blocks not eno…
rainyfly Jan 8, 2026
d05f5f0
[Cherry-Pick][Bugfix] Fix mtp logprob hang problem when include stop_…
Deleter-D Jan 8, 2026
9e542e8
[CI] Lock paddlepaddle-gpu/paddlepaddle-xpu==3.3.0 in release/2.4 (#5…
EmmonsCurse Jan 8, 2026
16645c6
[BugFix] fix xpu import set_data_ipc (#5945)
liyonghua0910 Jan 8, 2026
8049a49
[Cherry-Pick][Bugfix] Fix entropy calculation bugs (#5941) (#5942)
Deleter-D Jan 8, 2026
37bed64
[Cherry-Pick][BugFix] Fix misleading logging in worker_process for re…
Copilot Jan 9, 2026
741a015
[BugFix][Cherry-Pick] cp fix dyc8 cache bug(#5958) (#5959)
kevincheng2 Jan 9, 2026
f12b7a7
support_lastnorm_gather_split_r2.4 (#5925)
xiaoluomi Jan 9, 2026
bdaabf0
[Cherry-Pick][Speculative Decoding] Return accepted tokens per head i…
Deleter-D Jan 9, 2026
3e69022
[CI] Align PaddlePaddle version to latest due to tag change (#5971)
EmmonsCurse Jan 9, 2026
bbe9731
2.4_fix_mtp_forward_meta (#5977)
xiaoluomi Jan 9, 2026
0dfba18
[Cherry-Pick][BugFix] Fix entropy calculation issue in TP (#5997) #5998
Deleter-D Jan 13, 2026
72419c7
[Cherry-Pick][V1 Loader] Load safetensors weights in natural key orde…
bukejiyu Jan 13, 2026
491f692
[Cherry-Pick][BugFix] cp fix metrics cache tokens(#6001) (#6002)
kevincheng2 Jan 13, 2026
df197b2
[Cherry-Pick] [BugFix] fix cache transfer manager updating/clearing (…
liyonghua0910 Jan 13, 2026
303580d
[Cherry-Pick] [BugFix] Rename need_block_num_signal to fix shm name c…
liyonghua0910 Jan 14, 2026
8ce2623
Revert "[CI] Align PaddlePaddle version to latest due to tag change (…
EmmonsCurse Jan 14, 2026
b9b9c51
[Cherry-Pick][RL][CI] Support Async R3 And Add Accuracy Test #5937 (#…
gongshaotian Jan 15, 2026
caa186f
[Cherry-Pick][CI] Add commit-level Linux build task for RL(#5857) (#6…
EmmonsCurse Jan 16, 2026
a971755
[Cherry-Pick][BugFix] cp fix revert bug(#6061) (#6063)
kevincheng2 Jan 16, 2026
a512444
[Cherry-Pick][Bugfix] Fix MTP logprob issues caused by max_num_logpro…
Deleter-D Jan 19, 2026
a0f6950
[Cherry-Pick][XPU]XPU Release/2.4 Note(#6125)
iosmers Jan 20, 2026
fcf91c1
fix to_dict (#6138)
Deleter-D Jan 21, 2026
100b803
[Cherry-Pick] [RL] [APIServer] add more status codes for update/clear…
liyonghua0910 Jan 21, 2026
2977a28
[Cherry-Pick][Docs] Update docs for 2.4.0 tag (#6145) (#6147)
yangjianfengo1 Jan 21, 2026
7ba13f9
update requirements paddleformers 0.4.1 (#6136)
a31413510 Jan 22, 2026
9a91a5c
[Cherry-Pick][CI] Update image used by build_rl in ce_job.yml and fix…
EmmonsCurse Jan 22, 2026
122eae1
[WIP] origin code
gongshaotian Jan 22, 2026
5f04bd0
success save routing to cpu cache
gongshaotian Jan 23, 2026
29d5b89
delete self.seq_lens_decoder
gongshaotian Jan 23, 2026
3032632
[Cherry-Pick] [BugFix] move cache creation back to cache transfer pro…
liyonghua0910 Jan 23, 2026
739c8a2
[Cherry-Pick][Others] enable use PFCC/PaddleFleet deep_ep (#6191)
yuanlehome Jan 23, 2026
36ef77f
have put table bug
gongshaotian Jan 23, 2026
3868c2c
success put use stop flags
gongshaotian Jan 23, 2026
3c88a52
pass cpu_cache、block_table equal check
gongshaotian Jan 23, 2026
13ad267
[fix] fix cache config attribute error (#6199)
liyonghua0910 Jan 24, 2026
9a48206
[Feature] Unify quant ops (#6021)
fxyfxy777 Jan 24, 2026
f61a157
succes r3+prefixcache
gongshaotian Jan 23, 2026
65fc827
Fix chunk prefill and prefix cache task bug
gongshaotian Jan 24, 2026
da97b18
add long test case
gongshaotian Jan 24, 2026
64fb65e
Merge branch 'release/2.4' of https://github.com/PaddlePaddle/FastDep…
gongshaotian Jan 24, 2026
da9b356
Revert "[Feature] Unify quant ops (#6021)"
gongshaotian Jan 24, 2026
25a1d67
Fd use pfcc deepep or paddlefleet/deepep (#6206)
yuanlehome Jan 25, 2026
110dcad
fix initialize and acc bug
gongshaotian Jan 25, 2026
0321695
fix get position bug
gongshaotian Jan 25, 2026
c233798
refine code
gongshaotian Jan 25, 2026
5e7ab50
delete block table utils
gongshaotian Jan 25, 2026
1f55062
refine model runner code
gongshaotian Jan 25, 2026
1c01c9b
[BugFix] fix cache transfer tasks failure after cache cleared (#6201)
liyonghua0910 Jan 26, 2026
3c99a5d
Update _build_linux_rl.yml (#6215)
ZhangYulongg Jan 26, 2026
d8921a5
support routing pad to max_model_len
gongshaotian Jan 26, 2026
5cf2da4
[fix] fix pd_comm_port index out of bound (#6106)
liyonghua0910 Jan 26, 2026
c8cf686
[Cherry-Pick][Speculative Decoding] Support MTP for GLM-4.5-Air (#604…
Deleter-D Jan 27, 2026
1d519b9
[XPU][CI] Release ci update (#6212)
plusNew001 Jan 27, 2026
957bd2c
commit
zhoutianzi666 Jan 27, 2026
81d77d7
commit
zhoutianzi666 Jan 27, 2026
53f6fd4
add test tools
gongshaotian Jan 27, 2026
d0b94ec
Delete log and refine code
gongshaotian Jan 27, 2026
fb7ec62
[Cherry-Pick][Others] enhance deep_ep import and support mixed mode f…
yuanlehome Jan 27, 2026
c424287
[Cherry-Pick] update data_processor & add tool parser plugins#6096 (#…
luukunn Jan 27, 2026
22f0a5e
1.fix async numpy bug 2. refine code
gongshaotian Jan 27, 2026
d18d3fc
profile & refine code
gongshaotian Jan 27, 2026
7e4b3e3
add note
gongshaotian Jan 27, 2026
2855be9
Reapply "[Feature] Unify quant ops (#6021)"
gongshaotian Jan 27, 2026
fccfe57
Revert "[Cherry-Pick] update data_processor & add tool parser plugins…
Jiang-Jia-Jun Jan 27, 2026
80b54bf
Merge branch 'r3_prefixcache_2.4' into pr_6203
gongshaotian Jan 28, 2026
d259b56
Update Kernel #13 from zhoutianzi666/pr_6203
gongshaotian Jan 28, 2026
a4d4929
delete paremeter
gongshaotian Jan 28, 2026
9371b45
[Cherry-pick][RL] R3 Support Fused Put the Routing of All Layers#6099…
gongshaotian Jan 28, 2026
7b28ea7
fix transpose bug
gongshaotian Jan 28, 2026
4097455
[Cherry-Pick][RL] Support GLM MTP RL Model (#6223) (#6256)
Deleter-D Jan 28, 2026
fe80b01
fix insert decode task after set stop flag as true
gongshaotian Jan 28, 2026
f04ba48
[CI] Remove test_splitwise_scheduler and download latest_wheel explic…
EmmonsCurse Jan 29, 2026
fa7708b
merge 2.4
gongshaotian Jan 29, 2026
7bc4f05
Success Run StoreWrapper
gongshaotian Jan 31, 2026
2e37e94
WIP support async thread
gongshaotian Feb 2, 2026
a899fc2
success run async thread put
gongshaotian Feb 2, 2026
20733cd
fix clear prefix batch bug
gongshaotian Feb 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
66 changes: 43 additions & 23 deletions .github/workflows/_accuracy_test.yml
Original file line number Diff line number Diff line change
Expand Up @@ -39,29 +39,47 @@ jobs:
docker_image: ${{ inputs.DOCKER_IMAGE }}
fd_archive_url: ${{ inputs.FASTDEPLOY_ARCHIVE_URL }}
run: |
set -x
REPO="https://github.com/${{ github.repository }}.git"
FULL_REPO="${{ github.repository }}"
REPO_NAME="${FULL_REPO##*/}"
BASE_BRANCH="${{ github.base_ref }}"
docker pull ${docker_image}
# Clean the repository directory before starting
docker run --rm --net=host -v $(pwd):/workspace -w /workspace \
-e "REPO_NAME=${REPO_NAME}" \
${docker_image} /bin/bash -c '
if [ -d ${REPO_NAME} ]; then
echo "Directory ${REPO_NAME} exists, removing it..."
rm -rf ${REPO_NAME}*
set -x
REPO="https://github.com/${{ github.repository }}.git"
FULL_REPO="${{ github.repository }}"
REPO_NAME="${FULL_REPO##*/}"
BASE_BRANCH="${{ github.base_ref }}"
docker pull ${docker_image}
# Clean the repository directory before starting
docker run --rm --net=host -v $(pwd):/workspace -w /workspace \
-e "REPO_NAME=${REPO_NAME}" \
${docker_image} /bin/bash -c '
CLEAN_RETRIES=3
CLEAN_COUNT=0

while [ $CLEAN_COUNT -lt $CLEAN_RETRIES ]; do
echo "Attempt $((CLEAN_COUNT+1)) to remove ${REPO_NAME}* ..."
rm -rf "${REPO_NAME}"* || true
sleep 2

# Check if anything matching ${REPO_NAME}* still exists
if ! ls "${REPO_NAME}"* >/dev/null 2>&1; then
echo "All ${REPO_NAME}* removed successfully"
break
fi
'

wget -q --no-proxy ${fd_archive_url}
tar -xf FastDeploy.tar.gz
rm -rf FastDeploy.tar.gz
cd FastDeploy
git config --global user.name "FastDeployCI"
git config --global user.email "fastdeploy_ci@example.com"
git log -n 3 --oneline
CLEAN_COUNT=$((CLEAN_COUNT + 1))
done

if ls "${REPO_NAME}"* >/dev/null 2>&1; then
echo "ERROR: Failed to clean ${REPO_NAME}* after multiple attempts"
ls -ld "${REPO_NAME}"*
exit 1
fi
'

wget -q --no-proxy ${fd_archive_url}
tar -xf FastDeploy.tar.gz
rm -rf FastDeploy.tar.gz
cd FastDeploy
git config --global user.name "FastDeployCI"
git config --global user.email "fastdeploy_ci@example.com"
git log -n 3 --oneline

- name: Run FastDeploy Base Tests
shell: bash
Expand Down Expand Up @@ -143,14 +161,16 @@ jobs:
-v "${CACHE_DIR}/ConfigDir:/root/.config" \
-e TZ="Asia/Shanghai" \
--gpus '"device='"${DEVICES}"'"' ${docker_image} /bin/bash -xc '
python -m pip install --pre paddlepaddle-gpu -i https://www.paddlepaddle.org.cn/packages/nightly/cu126/
# Avoid using pip cache to ensure the wheel is updated to the latest version
wget -q --no-proxy https://paddle-qa.bj.bcebos.com/paddle-pipeline/Release-TagBuild-Training-Linux-Gpu-Cuda12.6-Cudnn9.5-Trt10.5-Mkl-Avx-Gcc11-SelfBuiltPypiUse/latest/paddlepaddle_gpu-0.0.0-cp310-cp310-linux_x86_64.whl
python -m pip install paddlepaddle_gpu-0.0.0-cp310-cp310-linux_x86_64.whl

pip config set global.index-url https://mirrors.tuna.tsinghua.edu.cn/pypi/web/simple

python -m pip install ${fastdeploy_wheel_url}
python -m pip install pytest

wget https://paddle-qa.bj.bcebos.com/zhengtianyu/tools/llm-deploy-linux-amd64
wget --no-proxy https://paddle-qa.bj.bcebos.com/zhengtianyu/tools/llm-deploy-linux-amd64
chmod +x ./llm-deploy-linux-amd64
./llm-deploy-linux-amd64 -python python3.10 \
-model_name ERNIE-4.5-0.3B-Paddle \
Expand Down
89 changes: 67 additions & 22 deletions .github/workflows/_base_test.yml
Original file line number Diff line number Diff line change
Expand Up @@ -39,29 +39,72 @@ jobs:
docker_image: ${{ inputs.DOCKER_IMAGE }}
fd_archive_url: ${{ inputs.FASTDEPLOY_ARCHIVE_URL }}
run: |
set -x
REPO="https://github.com/${{ github.repository }}.git"
FULL_REPO="${{ github.repository }}"
REPO_NAME="${FULL_REPO##*/}"
BASE_BRANCH="${{ github.base_ref }}"
docker pull ${docker_image}
# Clean the repository directory before starting
docker run --rm --net=host -v $(pwd):/workspace -w /workspace \
-e "REPO_NAME=${REPO_NAME}" \
${docker_image} /bin/bash -c '
if [ -d ${REPO_NAME} ]; then
echo "Directory ${REPO_NAME} exists, removing it..."
rm -rf ${REPO_NAME}*
set -x
REPO="https://github.com/${{ github.repository }}.git"
FULL_REPO="${{ github.repository }}"
REPO_NAME="${FULL_REPO##*/}"
BASE_BRANCH="${{ github.base_ref }}"
docker pull ${docker_image}
# Clean the repository directory before starting
docker run --rm --net=host -v $(pwd):/workspace -w /workspace \
-e "REPO_NAME=${REPO_NAME}" \
${docker_image} /bin/bash -c '
CLEAN_RETRIES=3
CLEAN_COUNT=0

while [ $CLEAN_COUNT -lt $CLEAN_RETRIES ]; do
echo "Attempt $((CLEAN_COUNT+1)) to remove ${REPO_NAME}* ..."
rm -rf "${REPO_NAME}"* || true
sleep 2

# Check if anything matching ${REPO_NAME}* still exists
if ! ls "${REPO_NAME}"* >/dev/null 2>&1; then
echo "All ${REPO_NAME}* removed successfully"
break
fi

CLEAN_COUNT=$((CLEAN_COUNT + 1))
done

if ls "${REPO_NAME}"* >/dev/null 2>&1; then
echo "ERROR: Failed to clean ${REPO_NAME}* after multiple attempts"
ls -ld "${REPO_NAME}"*
exit 1
fi
'

wget -q --no-proxy ${fd_archive_url}
tar -xf FastDeploy.tar.gz
rm -rf FastDeploy.tar.gz
cd FastDeploy
git config --global user.name "FastDeployCI"
git config --global user.email "fastdeploy_ci@example.com"
git log -n 3 --oneline
# Download with retry and validation
MAX_RETRIES=3
RETRY_COUNT=0
while [ $RETRY_COUNT -lt $MAX_RETRIES ]; do
if wget -q --no-proxy ${fd_archive_url} && [ -f FastDeploy.tar.gz ] && [ -s FastDeploy.tar.gz ]; then
echo "Download successful, file size: $(stat -c%s FastDeploy.tar.gz) bytes"
break
else
RETRY_COUNT=$((RETRY_COUNT + 1))
echo "Download failed or file is empty, retry $RETRY_COUNT/$MAX_RETRIES..."
rm -f FastDeploy.tar.gz
sleep 2
fi
done

if [ ! -f FastDeploy.tar.gz ] || [ ! -s FastDeploy.tar.gz ]; then
echo "ERROR: Failed to download FastDeploy.tar.gz after $MAX_RETRIES attempts"
exit 1
fi

# Verify tar.gz integrity before extraction
if ! tar -tzf FastDeploy.tar.gz > /dev/null 2>&1; then
echo "ERROR: FastDeploy.tar.gz is corrupted or incomplete"
exit 1
fi

tar -xf FastDeploy.tar.gz
rm -rf FastDeploy.tar.gz
cd FastDeploy
git config --global user.name "FastDeployCI"
git config --global user.email "fastdeploy_ci@example.com"
git log -n 3 --oneline

- name: Run FastDeploy Base Tests
shell: bash
Expand Down Expand Up @@ -143,7 +186,9 @@ jobs:
-v "${CACHE_DIR}/ConfigDir:/root/.config" \
-e TZ="Asia/Shanghai" \
--gpus '"device='"${DEVICES}"'"' ${docker_image} /bin/bash -xc '
python -m pip install --pre paddlepaddle-gpu -i https://www.paddlepaddle.org.cn/packages/nightly/cu126/
# Avoid using pip cache to ensure the wheel is updated to the latest version
wget -q --no-proxy https://paddle-qa.bj.bcebos.com/paddle-pipeline/Release-TagBuild-Training-Linux-Gpu-Cuda12.6-Cudnn9.5-Trt10.5-Mkl-Avx-Gcc11-SelfBuiltPypiUse/latest/paddlepaddle_gpu-0.0.0-cp310-cp310-linux_x86_64.whl
python -m pip install paddlepaddle_gpu-0.0.0-cp310-cp310-linux_x86_64.whl

pip config set global.index-url https://mirrors.tuna.tsinghua.edu.cn/pypi/web/simple

Expand Down Expand Up @@ -209,7 +254,7 @@ jobs:
export TEMPLATE=TOKEN_NORMAL
curl -X POST http://0.0.0.0:${FLASK_PORT}/switch \
-H "Content-Type: application/json" \
-d "{\"--model\": \"/MODELDATA/ERNIE-4.5-VL-28B-A3B-Thinking\", \"--reasoning-parser\": \"ernie-45-vl-thinking\", \"--tool-call-parser\": \"ernie-45-vl-thinking\", \"--tensor-parallel-size\": 1, \"--quantization\": \"wint4\", \"--max-model-len\": 131072, \"--max-num-seqs\": 32}"
-d "{\"--model\": \"/MODELDATA/ERNIE-4.5-VL-28B-A3B-Thinking\", \"--reasoning-parser\": \"ernie-45-vl-thinking\", \"--tool-call-parser\": \"ernie-45-vl-thinking\", \"--tensor-parallel-size\": 1, \"--quantization\": \"wint4\", \"--max-model-len\": 131072, \"--max-num-seqs\": 32, \"--no-enable-prefix-caching\": true}"
check_service 90
python -m pytest -sv test_prompt_ids.py || TEST_EXIT_CODE=1

Expand Down
28 changes: 24 additions & 4 deletions .github/workflows/_build_linux.yml
Original file line number Diff line number Diff line change
Expand Up @@ -76,9 +76,27 @@ jobs:
docker run --rm --net=host -v $(pwd):/workspace -w /workspace \
-e "REPO_NAME=${REPO_NAME}" \
${docker_image} /bin/bash -c '
if [ -d ${REPO_NAME} ]; then
echo "Directory ${REPO_NAME} exists, removing it..."
rm -rf ${REPO_NAME}*
CLEAN_RETRIES=3
CLEAN_COUNT=0

while [ $CLEAN_COUNT -lt $CLEAN_RETRIES ]; do
echo "Attempt $((CLEAN_COUNT+1)) to remove ${REPO_NAME}* ..."
rm -rf "${REPO_NAME}"* || true
sleep 2

# Check if anything matching ${REPO_NAME}* still exists
if ! ls "${REPO_NAME}"* >/dev/null 2>&1; then
echo "All ${REPO_NAME}* removed successfully"
break
fi

CLEAN_COUNT=$((CLEAN_COUNT + 1))
done

if ls "${REPO_NAME}"* >/dev/null 2>&1; then
echo "ERROR: Failed to clean ${REPO_NAME}* after multiple attempts"
ls -ld "${REPO_NAME}"*
exit 1
fi
'

Expand Down Expand Up @@ -155,7 +173,9 @@ jobs:
elif [[ "${PADDLEVERSION}" != "" ]];then
python -m pip install paddlepaddle-gpu==${PADDLEVERSION} -i https://www.paddlepaddle.org.cn/packages/stable/cu126/
else
python -m pip install --pre paddlepaddle-gpu -i https://www.paddlepaddle.org.cn/packages/nightly/cu126/
# Avoid using pip cache to ensure the wheel is updated to the latest version
wget -q --no-proxy https://paddle-qa.bj.bcebos.com/paddle-pipeline/Release-TagBuild-Training-Linux-Gpu-Cuda12.6-Cudnn9.5-Trt10.5-Mkl-Avx-Gcc11-SelfBuiltPypiUse/latest/paddlepaddle_gpu-0.0.0-cp310-cp310-linux_x86_64.whl
python -m pip install paddlepaddle_gpu-0.0.0-cp310-cp310-linux_x86_64.whl
fi

pip config set global.index-url https://mirrors.tuna.tsinghua.edu.cn/pypi/web/simple
Expand Down
Loading
Loading