You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
What is the bug?
Model zip file for ONNX Llama 3.2 1b Instruct is about 2.8GB, which is under 4GB limit. Register model call succeeds, but the registration task never gets to COMPLETED state.
The model zip file is successfully downloaded to data/ml_cache/models_cache/register/task-id/version/llama-3.2-1b_onnx.zip. The chunking of the model file casts model file size to int making it a negative number in FileUtils. This results in an empty list of chunks.
How can one reproduce the bug?
Use model zip file greater than 2GB.
What is the expected behavior?
A clear and concise description of what you expected to happen.
What is your host/environment?
OS: Ubuntu 24.04
Do you have any screenshots?
N/A
Do you have any additional context?
Log shows download succeeded and not any other info about the task:
[2024-11-03T13:08:59,414][INFO ][o.o.m.m.MLModelManager ] [user-laptop] create new model meta doc lw-l85IBso2zKYtHJdUD for register model task lg-l85IBso2zKYtHJNW_
[2024-11-03T13:08:59,415][DEBUG][o.o.m.e.ModelHelper ] [user-laptop] download model to file /home/user/search/opensearch/opensearch-2.15.0/data/ml_cache/models_cache/register/lw-l85IBso2zKYtHJdUD/6/llama-3.2-1b_onnx.zip
Downloading: 0% |█ | llama-3.2-1b_onnx.zip
Downloading: 0% |█ | llama-3.2-1b_onnx.zipng: 0% |█ | llama-3.2-1b_onnx.zip
Downloading: 100% |████████████████████████████████████████| llama-3.2-1b_onnx.zipg: 0% |█ | llama-3.2-1b_onnx.zip
[2024-11-03T13:10:18,881][INFO ][o.o.j.s.JobSweeper ] [user-laptop] Running full sweep
[2024-11-03T13:15:18,882][INFO ][o.o.j.s.JobSweeper ] [user-laptop] Running full sweep
The task remains in CREATED state indefinitely.
The text was updated successfully, but these errors were encountered:
What is the bug?
Model zip file for ONNX Llama 3.2 1b Instruct is about 2.8GB, which is under 4GB limit. Register model call succeeds, but the registration task never gets to COMPLETED state.
The model zip file is successfully downloaded to data/ml_cache/models_cache/register/task-id/version/llama-3.2-1b_onnx.zip. The chunking of the model file casts model file size to int making it a negative number in FileUtils. This results in an empty list of chunks.
How can one reproduce the bug?
Use model zip file greater than 2GB.
What is the expected behavior?
A clear and concise description of what you expected to happen.
What is your host/environment?
Do you have any screenshots?
N/A
Do you have any additional context?
Log shows download succeeded and not any other info about the task:
The task remains in CREATED state indefinitely.
The text was updated successfully, but these errors were encountered: