Dear team,

I am novice here and trying to look some options available in ClearML for our usage..

I am facing an issue in "New Run" of a successfully executed Pipeline (ID: 2b68e367282a4e6b99d2312718ceb3df ) in ClearML.
To give you more details,

• The pipeline is triggered from my personal mac machine with all the dependencies (requirements.txt) installed in venv.
• The clearml-agent is running in my machine with this command "clearml-agent daemon --queue default"
• I am trying to trigger new run from ClearML and getting the pytorch==2.8.0 wheel installation issue (ID: a7a5028ebdb04ec3b0140373cac9b79d ), however, all this worked from my machine.
  Error :

learml_agent: Warning: could not resolve python wheel replacement for torch==2.8.0
clearml_agent: ERROR: Could not install task requirements!
Exception when trying to resolve python wheel: Could not find pytorch wheel URL for: torch==2.8.0 with cuda 0 support

Instructions in clearml-conf file:

agent {
# Git configuration for SSH
git_user: ""
git_pass: ""

# This forces SSH protocol
force_git_ssh_protocol: true

# Default Python version
default_python: "3.9"

# Package manager
package_manager {
    type: pip
}

# Working directory
force_git_root_python_path: true

}

venvs_cache {

Use cached virtual environments for tasks with identical dependencies

enabled = yes
}

Instructions in PipelineController:

""Create and configure the ClearML pipeline."""
pipe = PipelineController(
name="RAI Validator - Bias Detection Model Training Pipeline - New(2)",
project="Disseqt_BIAS_Validator_Model_Builder(3)",
version="1.0",
add_pipeline_tags=True,
repo="git@github.com:DisseqtAI/ml-ai-model-development-training-inference.git", # SSH format
repo_branch="main",
working_dir="clearml_pipelines",
abort_on_failure=True
)

and one of the pipeline step, others follow the same type of instruction:

pipe.add_function_step(
name="preprocess_data",
function=preprocess_data,
function_kwargs={
"input_data_path": "${pipeline.input_data_path}",
"output_data_path": "${pipeline.output_data_path}",
"model_type": "${pipeline.model_choice}"
},
function_return=["processed_data_path"],
# packages=packages,
packages="./requirements.txt",
cache_executed_step=False
)

Kindly do let me know, what could be the issue ? and what is that I am missing here ?