Exception when using torchserve to deploy hugging face model: java.lang.InterruptedException: null · pytorch/serve#3026

(4 comments) (0 reactions) (0 assignees)Java (3,844 stars) (790 forks)batch import

help wantedneeds-reproductiontriaged

説明

🐛 Describe the bug

I followed the tutorial as https://github.com/pytorch/serve/tree/master/examples/Huggingface_Transformers

First,

python Download_Transformer_models.py

Then,

torch-model-archiver --model-name BERTSeqClassification --version 1.0 --serialized-file Transformer_model/pytorch_model.bin --handler ./Transformer_handler_generalized.py --extra-files "Transformer_model/config.json,./setup_config.json,./Seq_classification_artifacts/index_to_name.json"

Finally,

 torchserve --start --model-store model_store --models my_tc=BERTSeqClassification.mar --ncs

The system cannot start as usualy, it gives out the error log, throwing an Exception

java.lang.InterruptedException: null
        at java.util.concurrent.locks.AbstractQueuedSynchronizer$ConditionObject.awaitNanos(AbstractQueuedSynchronizer.java:1679) ~[?:?]
        at java.util.concurrent.LinkedBlockingDeque.pollFirst(LinkedBlockingDeque.java:515) ~[?:?]
        at java.util.concurrent.LinkedBlockingDeque.poll(LinkedBlockingDeque.java:677) ~[?:?]
        at org.pytorch.serve.wlm.Model.pollBatch(Model.java:367) ~[model-server.jar:?]
        at org.pytorch.serve.wlm.BatchAggregator.getRequest(BatchAggregator.java:36) ~[model-server.jar:?]
        at org.pytorch.serve.wlm.WorkerThread.run(WorkerThread.java:194) [model-server.jar:?]
        at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1136) [?:?]
        at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:635) [?:?]
        at java.lang.Thread.run(Thread.java:833) [?:?]

I tried curl to check the model

root@0510f3693f42:/home/model-server# curl  http://127.0.0.1:8081/models           
{
  "models": []
}

Error logs

2024-03-14T07:34:24,938 [INFO ] epollEventLoopGroup-5-17 org.pytorch.serve.wlm.WorkerThread - 9015 Worker disconnected. WORKER_STARTED 2024-03-14T07:34:24,938 [INFO ] W-9015-my_tc_1.0-stdout MODEL_LOG - Connection accepted: /home/model-server/tmp/.ts.sock.9015. 2024-03-14T07:34:24,938 [DEBUG] W-9015-my_tc_1.0 org.pytorch.serve.wlm.WorkerThread - System state is : WORKER_STARTED 2024-03-14T07:34:24,938 [DEBUG] W-9015-my_tc_1.0 org.pytorch.serve.wlm.WorkerThread - Backend worker monitoring thread interrupted or backend worker process died. java.lang.InterruptedException: null at java.util.concurrent.locks.AbstractQueuedSynchronizer$ConditionObject.awaitNanos(AbstractQueuedSynchronizer.java:1679) ~[?:?] at java.util.concurrent.LinkedBlockingDeque.pollFirst(LinkedBlockingDeque.java:515) ~[?:?] at java.util.concurrent.LinkedBlockingDeque.poll(LinkedBlockingDeque.java:677) ~[?:?] at org.pytorch.serve.wlm.Model.pollBatch(Model.java:367) ~[model-server.jar:?] at org.pytorch.serve.wlm.BatchAggregator.getRequest(BatchAggregator.java:36) ~[model-server.jar:?] at org.pytorch.serve.wlm.WorkerThread.run(WorkerThread.java:194) [model-server.jar:?] at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1136) [?:?] at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:635) [?:?] at java.lang.Thread.run(Thread.java:833) [?:?] 2024-03-14T07:34:24,938 [DEBUG] W-9015-my_tc_1.0 org.pytorch.serve.wlm.WorkerThread - W-9015-my_tc_1.0 State change WORKER_STARTED -> WORKER_STOPPED 2024-03-14T07:34:24,938 [WARN ] W-9015-my_tc_1.0 org.pytorch.serve.wlm.WorkerThread - Auto recovery failed again 2024-03-14T07:34:24,939 [WARN ] W-9015-my_tc_1.0 org.pytorch.serve.wlm.WorkerLifeCycle - terminateIOStreams() threadName=W-9015-my_tc_1.0-stderr 2024-03-14T07:34:24,939 [WARN ] W-9015-my_tc_1.0 org.pytorch.serve.wlm.WorkerLifeCycle - terminateIOStreams() threadName=W-9015-my_tc_1.0-stdout 2024-03-14T07:34:24,939 [INFO ] W-9015-my_tc_1.0 org.pytorch.serve.wlm.WorkerThread - Retry worker: 9015 in 3 seconds. 2024-03-14T07:34:24,946 [INFO ] W-9015-my_tc_1.0-stdout org.pytorch.serve.wlm.WorkerLifeCycle - Stopped Scanner - W-9015-my_tc_1.0-stdout 2024-03-14T07:34:24,946 [INFO ] W-9015-my_tc_1.0-stderr org.pytorch.serve.wlm.WorkerLifeCycle - Stopped Scanner - W-9015-my_tc_1.0-stderr 2024-03-14T07:34:27,207 [DEBUG] W-9010-my_tc_1.0 org.pytorch.serve.wlm.WorkerLifeCycle - Worker cmdline: [/home/venv/bin/python, /home/venv/lib/python3.9/site-packages/ts/model_service_worker.py, --sock-type, unix, --sock-name, /home/model-server/tmp/.ts.sock.9010, --metrics-config, /home/venv/lib/python3.9/site-packages/ts/configs/metrics.yaml] 2024-03-14T07:34:27,489 [DEBUG] W-9012-my_tc_1.0 org.pytorch.serve.wlm.WorkerLifeCycle - Worker cmdline: [/home/venv/bin/python, /home/venv/lib/python3.9/site-packages/ts/model_service_worker.py, --sock-type, unix, --sock-name, /home/model-server/tmp/.ts.sock.9012, --metrics-config, /home/venv/lib/python3.9/site-packages/ts/configs/metrics.yaml] 2024-03-14T07:34:27,579 [DEBUG] W-9000-my_tc_1.0 org.pytorch.serve.wlm.WorkerLifeCycle - Worker cmdline: [/home/venv/bin/python, /home/venv/lib/python3.9/site-packages/ts/model_service_worker.py, --sock-type, unix, --sock-name, /home/model-server/tmp/.ts.sock.9000, --metrics-config, /home/venv/lib/python3.9/site-packages/ts/configs/metrics.yaml] 2024-03-14T07:34:27,669 [DEBUG] W-9011-my_tc_1.0 org.pytorch.serve.wlm.WorkerLifeCycle - Worker cmdline: [/home/venv/bin/python, /home/venv/lib/python3.9/site-packages/ts/model_service_worker.py, --sock-type, unix, --sock-name, /home/model-server/tmp/.ts.sock.9011, --metrics-config, /home/venv/lib/python3.9/site-packages/ts/configs/metrics.yaml] 2024-03-14T07:34:27,704 [DEBUG] W-9006-my_tc_1.0 org.pytorch.serve.wlm.WorkerLifeCycle - Worker cmdline: [/home/venv/bin/python, /home/venv/lib/python3.9/site-packages/ts/model_service_worker.py, --sock-type, unix, --sock-name, /home/model-server/tmp/.ts.sock.9006, --metrics-config, /home/venv/lib/python3.9/site-packages/ts/configs/metrics.yaml] 2024-03-14T07:34:27,707 [DEBUG] W-9008-my_tc_1.0 org.pytorch.serve.wlm.WorkerLifeCycle - Worker cmdline: [/home/venv/bin/python, /home/venv/lib/python3.9/site-packages/ts/model_service_worker.py, --sock-type, unix, --sock-name, /home/model-server/tmp/.ts.sock.9008, --metrics-config, /home/venv/lib/python3.9/site-packages/ts/configs/metrics.yaml] 2024-03-14T07:34:27,734 [DEBUG] W-9017-my_tc_1.0 org.pytorch.serve.wlm.WorkerLifeCycle - Worker cmdline: [/home/venv/bin/python, /home/venv/lib/python3.9/site-packages/ts/model_service_worker.py, --sock-type, unix, --sock-name, /home/model-server/tmp/.ts.sock.9017, --metrics-config, /home/venv/lib/python3.9/site-packages/ts/configs/metrics.yaml] 2024-03-14T07:34:27,751 [DEBUG] W-9003-my_tc_1.0 org.pytorch.serve.wlm.WorkerLifeCycle - Worker cmdline: [/home/venv/bin/python, /home/venv/lib/python3.9/site-packages/ts/model_service_worker.py, --sock-type, unix, --sock-name, /home/model-server/tmp/.ts.sock.9003, --metrics-config, /home/venv/lib/python3.9/site-packages/ts/configs/metrics.yaml] 2024-03-14T07:34:27,776 [DEBUG] W-9001-my_tc_1.0 org.pytorch.serve.wlm.WorkerLifeCycle - Worker cmdline: [/home/venv/bin/python, /home/venv/lib/python3.9/site-packages/ts/model_service_worker.py, --sock-type, unix, --sock-name, /home/model-server/tmp/.ts.sock.9001, --metrics-config, /home/venv/lib/python3.9/site-packages/ts/configs/metrics.yaml] 2024-03-14T07:34:27,804 [DEBUG] W-9005-my_tc_1.0 org.pytorch.serve.wlm.WorkerLifeCycle - Worker cmdline: [/home/venv/bin/python, /home/venv/lib/python3.9/site-packages/ts/model_service_worker.py, --sock-type, unix, --sock-name, /home/model-server/tmp/.ts.sock.9005, --metrics-config, /home/venv/lib/python3.9/site-packages/ts/configs/metrics.yaml] 2024-03-14T07:34:27,815 [DEBUG] W-9009-my_tc_1.0 org.pytorch.serve.wlm.WorkerLifeCycle - Worker cmdline: [/home/venv/bin/python, /home/venv/lib/python3.9/site-packages/ts/model_service_worker.py, --sock-type, unix, --sock-name, /home/model-server/tmp/.ts.sock.9009, --metrics-config, /home/venv/lib/python3.9/site-packages/ts/configs/metrics.yaml] 2024-03-14T07:34:27,844 [DEBUG] W-9013-my_tc_1.0 org.pytorch.serve.wlm.WorkerLifeCycle - Worker cmdline: [/home/venv/bin/python, /home/venv/lib/python3.9/site-packages/ts/model_service_worker.py, --sock-type, unix, --sock-name, /home/model-server/tmp/.ts.sock.9013, --metrics-config, /home/venv/lib/python3.9/site-packages/ts/configs/metrics.yaml] 2024-03-14T07:34:27,848 [DEBUG] W-9004-my_tc_1.0 org.pytorch.serve.wlm.WorkerLifeCycle - Worker cmdline: [/home/venv/bin/python, /home/venv/lib/python3.9/site-packages/ts/model_service_worker.py, --sock-type, unix, --sock-name, /home/model-server/tmp/.ts.sock.9004, --metrics-config, /home/venv/lib/python3.9/site-packages/ts/configs/metrics.yaml] 2024-03-14T07:34:27,853 [DEBUG] W-9007-my_tc_1.0 org.pytorch.serve.wlm.WorkerLifeCycle - Worker cmdline: [/home/venv/bin/python, /home/venv/lib/python3.9/site-packages/ts/model_service_worker.py, --sock-type, unix, --sock-name, /home/model-server/tmp/.ts.sock.9007, --metrics-config, /home/venv/lib/python3.9/site-packages/ts/configs/metrics.yaml] 2024-03-14T07:34:27,857 [DEBUG] W-9019-my_tc_1.0 org.pytorch.serve.wlm.WorkerLifeCycle - Worker cmdline: [/home/venv/bin/python, /home/venv/lib/python3.9/site-packages/ts/model_service_worker.py, --sock-type, unix, --sock-name, /home/model-server/tmp/.ts.sock.9019, --metrics-config, /home/venv/lib/python3.9/site-packages/ts/configs/metrics.yaml] 2024-03-14T07:34:27,871 [DEBUG] W-9002-my_tc_1.0 org.pytorch.serve.wlm.WorkerLifeCycle - Worker cmdline: [/home/venv/bin/python, /home/venv/lib/python3.9/site-packages/ts/model_service_worker.py, --sock-type, unix, --sock-name, /home/model-server/tmp/.ts.sock.9002, --metrics-config, /home/venv/lib/python3.9/site-packages/ts/configs/metrics.yaml] 2024-03-14T07:34:27,904 [DEBUG] W-9014-my_tc_1.0 org.pytorch.serve.wlm.WorkerLifeCycle - Worker cmdline: [/home/venv/bin/python, /home/venv/lib/python3.9/site-packages/ts/model_service_worker.py, --sock-type, unix, --sock-name, /home/model-server/tmp/.ts.sock.9014, --metrics-config, /home/venv/lib/python3.9/site-packages/ts/configs/metrics.yaml] 2024-03-14T07:34:27,904 [DEBUG] W-9018-my_tc_1.0 org.pytorch.serve.wlm.WorkerLifeCycle - Worker cmdline: [/home/venv/bin/python, /home/venv/lib/python3.9/site-packages/ts/model_service_worker.py, --sock-type, unix, --sock-name, /home/model-server/tmp/.ts.sock.9018, --metrics-config, /home/venv/lib/python3.9/site-packages/ts/configs/metrics.yaml] 2024-03-14T07:34:27,927 [DEBUG] W-9016-my_tc_1.0 org.pytorch.serve.wlm.WorkerLifeCycle - Worker cmdline: [/home/venv/bin/python, /home/venv/lib/python3.9/site-packages/ts/model_service_worker.py, --sock-type, unix, --sock-name, /home/model-server/tmp/.ts.sock.9016, --metrics-config, /home/venv/lib/python3.9/site-packages/ts/configs/metrics.yaml] 2024-03-14T07:34:27,939 [DEBUG] W-9015-my_tc_1.0 org.pytorch.serve.wlm.WorkerLifeCycle - Worker cmdline: [/home/venv/bin/python, /home/venv/lib/python3.9/site-packages/ts/model_service_worker.py, --sock-type, unix, --sock-name, /home/model-server/tmp/.ts.sock.9015, --metrics-config, /home/venv/lib/python3.9/site-packages/ts/configs/metrics.yaml] 2024-03-14T07:34:28,642 [INFO ] W-9010-my_tc_1.0-stdout MODEL_LOG - s_name_part0=/home/model-server/tmp/.ts.sock, s_name_part1=9010, pid=8906 2024-03-14T07:34:28,644 [INFO ] W-9010-my_tc_1.0-stdout MODEL_LOG - Listening on port: /home/model-server/tmp/.ts.sock.9010 2024-03-14T07:34:28,657 [INFO ] W-9010-my_tc_1.0-stdout MODEL_LOG - Successfully loaded /home/venv/lib/python3.9/site-packages/ts/configs/metrics.yaml. 2024-03-14T07:34:28,658 [INFO ] W-9010-my_tc_1.0-stdout MODEL_LOG - [PID]8906 2024-03-14T07:34:28,658 [INFO ] W-9010-my_tc_1.0-stdout MODEL_LOG - Torch worker started. 2024-03-14T07:34:28,659 [DEBUG] W-9010-my_tc_1.0 org.pytorch.serve.wlm.WorkerThread - W-9010-my_tc_1.0 State change WORKER_STOPPED -> WORKER_STARTED 2024-03-14T07:34:28,659 [INFO ] W-9010-my_tc_1.0 org.pytorch.serve.wlm.WorkerThread - Connecting to: /home/model-server/tmp/.ts.sock.9010 2024-03-14T07:34:28,660 [INFO ] epollEventLoopGroup-5-6 org.pytorch.serve.wlm.WorkerThread - 9010 Worker disconnected. WORKER_STARTED 2024-03-14T07:34:28,661 [DEBUG] W-9010-my_tc_1.0 org.pytorch.serve.wlm.WorkerThread - System state is : WORKER_STARTED 2024-03-14T07:34:28,661 [DEBUG] W-9010-my_tc_1.0 org.pytorch.serve.wlm.WorkerThread - Backend worker monitoring thread interrupted or backend worker process died. java.lang.InterruptedException: null at java.util.concurrent.locks.AbstractQueuedSynchronizer.tryAcquireSharedNanos(AbstractQueuedSynchronizer.java:1081) ~[?:?] at java.util.concurrent.CountDownLatch.await(CountDownLatch.java:276) ~[?:?] at org.pytorch.serve.wlm.WorkerThread.connect(WorkerThread.java:424) ~[model-server.jar:?] at org.pytorch.serve.wlm.WorkerThread.run(WorkerThread.java:191) [model-server.jar:?] at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1136) [?:?] at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:635) [?:?] at java.lang.Thread.run(Thread.java:833) [?:?] 2024-03-14T07:34:28,661 [DEBUG] W-9010-my_tc_1.0 org.pytorch.serve.wlm.WorkerThread - W-9010-my_tc_1.0 State change WORKER_STARTED -> WORKER_STOPPED 2024-03-14T07:34:28,662 [WARN ] W-9010-my_tc_1.0 org.pytorch.serve.wlm.WorkerThread - Auto recovery failed again 2024-03-14T07:34:28,663 [WARN ] W-9010-my_tc_1.0 org.pytorch.serve.wlm.WorkerLifeCycle - terminateIOStreams() threadName=W-9010-my_tc_1.0-stderr 2024-03-14T07:34:28,663 [WARN ] W-9010-my_tc_1.0 org.pytorch.serve.wlm.WorkerLifeCycle - terminateIOStreams() threadName=W-9010-my_tc_1.0-stdout 2024-03-14T07:34:28,664 [INFO ] W-9010-my_tc_1.0 org.pytorch.serve.wlm.WorkerThread - Retry worker: 9010 in 5 seconds. 2024-03-14T07:34:28,692 [ERROR] epollEventLoopGroup-5-1 org.pytorch.serve.wlm.WorkerThread - Unknown exception io.netty.channel.unix.Errors$NativeIoException: readAddress(..) failed: Connection reset by peer 2024-03-14T07:34:28,698 [INFO ] W-9010-my_tc_1.0-stdout org.pytorch.serve.wlm.WorkerLifeCycle - Stopped Scanner - W-9010-my_tc_1.0-stdout 2024-03-14T07:34:28,698 [INFO ] W-9010-my_tc_1.0-stderr org.pytorch.serve.wlm.WorkerLifeCycle - Stopped Scanner - W-9010-my_tc_1.0-stderr

Installation instructions

pip install torchserve.

Yes, I am using docker image pytorch/torchserve:latest

Model Packaing

I use transformers=3.4.0 to save the pretrained model into

root@1796bda67dbf:~/Huggingface_Transformers# ll Transformer_model/
total 428008
drwxr-xr-x 2 root root      4096 Mar 14 07:05 ./
drwxr-xr-x 9 root root      4096 Mar 14 07:13 ../
-rw-r--r-- 1 root root       522 Mar 14 07:05 config.json
-rw-r--r-- 1 root root 438019213 Mar 14 07:05 pytorch_model.bin
-rw-r--r-- 1 root root       112 Mar 14 07:05 special_tokens_map.json
-rw-r--r-- 1 root root       174 Mar 14 07:05 tokenizer_config.json
-rw-r--r-- 1 root root    231508 Mar 14 07:05 vocab.txt

config.properties

No response

Versions

torchserve==0.9.0 torch-model-archiver==0.9.0

Python version: 3.9 (64-bit runtime) Python executable: /home/venv/bin/python

Versions of relevant python libraries: captum==0.6.0 numpy==1.24.3 psutil==5.9.5 requests==2.31.0 torch==2.1.0+cpu torch-model-archiver==0.9.0 torch-workflow-archiver==0.2.11 torchaudio==2.1.0+cpu torchdata==0.7.0 torchserve==0.9.0 torchtext==0.16.0+cpu torchvision==0.16.0+cpu wheel==0.40.0 torch==2.1.0+cpu torchtext==0.16.0+cpu torchvision==0.16.0+cpu torchaudio==2.1.0+cpu

Java Version:

OS: Ubuntu 20.04.6 LTS GCC version: (Ubuntu 9.4.0-1ubuntu1~20.04.2) 9.4.0 Clang version: N/A CMake version: N/A

Environment: library_path (LD_/DYLD_):

Repro instructions

As described above

Possible Solution

No response

コントリビューターガイド

技術スタック: pythonjava
領域: backendmachine learning
Issue 種別: bug
難度: 4
推定時間: 1-3 hours
活動状況: fresh
明確さ: needs investigation
前提条件: Familiarity with TorchServeBasic understanding of Java concurrencyAbility to run TorchServe locally
初心者向け度: 30
調査方針: The error occurs in Model.pollBatch (line 367 in Model.java) when it calls LinkedBlockingDeque.poll, which throws InterruptedException. Investigate why the worker thread is interrupted during polling. Check if there is a timeout or shutdown sequence that causes the interrupt. Also review the WorkerThread.run method and BatchAggregator.getRequest to understand the request flow. Look at the TorchServe configuration for batch settings and worker lifecycle. The logs show repeated worker start/stop cycles, suggesting a misconfiguration or bug in worker management.