Fixing database bug

[thaisacs]

Before this PR, when I tried to tune a model whose pipeline had 0 trials and the database directory didn't exist yet, I would get the following error:

[14:04:12] /home/thais/Dev/tvm/src/relax/transform/meta_schedule.cc:91: Warning: Creating JSONDatabase. Workload at: ./tuning_logs_alexnet/database_workload.json, Tuning records at: ./tuning_logs_alexnet/database_tuning_record.json
Traceback (most recent call last):
File "/home/thais/Dev/TVM-MetaSchedule-Bench/model-tune.py", line 108, in
tune_model(network_arg, args.dtype, TOTAL_TRIALS, target, work_dir)
File "/home/thais/Dev/TVM-MetaSchedule-Bench/model-tune.py", line 31, in tune_model
mod = relax.get_pipeline(
File "/home/thais/Dev/tvm/python/tvm/ir/transform.py", line 167, in call
return _ffi_transform_api.RunPass(self, mod)
File "python/tvm_ffi/cython/function.pxi", line 758, in core.Function.call
File "", line 0, in tvm::transform::Pass::operator()(tvm::IRModule) const
File "", line 0, in tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
File "", line 0, in tvm::transform::ModulePassNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
File "", line 0, in std::_Function_handler<tvm::IRModule (tvm::IRModule, tvm::transform::PassContext), tvm::transform::__TVMFFIStaticInitFunc4()::{lambda(tvm::ffi::TypedFunction<tvm::IRModule (tvm::ffi::RValueRef<tvm::IRModule, void>, tvm::transform::PassContext)>, tvm::transform::PassInfo)#1}::operator()(tvm::ffi::TypedFunction<tvm::IRModule (tvm::ffi::RValueRef<tvm::IRModule, void>, tvm::transform::PassContext)>, tvm::transform::PassInfo) const::{lambda(tvm::IRModule, tvm::transform::PassContext)#1}>::_M_invoke(std::_Any_data const&, tvm::IRModule&&, tvm::transform::PassContext&&)
File "", line 0, in tvm::transform::__TVMFFIStaticInitFunc4()::{lambda(tvm::ffi::TypedFunction<tvm::IRModule (tvm::ffi::RValueRef<tvm::IRModule, void>, tvm::transform::PassContext)>, tvm::transform::PassInfo)#1}::operator()(tvm::ffi::TypedFunction<tvm::IRModule (tvm::ffi::RValueRef<tvm::IRModule, void>, tvm::transform::PassContext)>, tvm::transform::PassInfo) const::{lambda(tvm::IRModule, tvm::transform::PassContext)#1}::operator()(tvm::IRModule, tvm::transform::PassContext) const
File "python/tvm_ffi/cython/function.pxi", line 914, in core.tvm_ffi_callback
File "/home/thais/Dev/tvm/python/tvm/relax/pipeline.py", line 172, in _pipeline
mod = tvm.transform.Sequential(
File "/home/thais/Dev/tvm/python/tvm/ir/transform.py", line 167, in call
return _ffi_transform_api.RunPass(self, mod)
File "python/tvm_ffi/cython/function.pxi", line 758, in core.Function.call
File "", line 0, in tvm::transform::Pass::operator()(tvm::IRModule) const
File "", line 0, in tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
File "", line 0, in tvm::transform::SequentialNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
File "", line 0, in tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
File "", line 0, in tvm::transform::ModulePassNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
File "", line 0, in std::_Function_handler<tvm::IRModule (tvm::IRModule, tvm::transform::PassContext), tvm::relax::transform::MetaScheduleApplyDatabase(tvm::ffi::Optional<tvm::ffi::String, void>, bool)::{lambda(tvm::IRModule, tvm::transform::PassContext)#1}>::_M_invoke(std::_Any_data const&, tvm::IRModule&&, tvm::transform::PassContext&&)
File "", line 0, in tvm::relax::transform::MetaScheduleApplyDatabase(tvm::ffi::Optional<tvm::ffi::String, void>, bool)::{lambda(tvm::IRModule, tvm::transform::PassContext)#1}::operator()(tvm::IRModule, tvm::transform::PassContext) const [clone .constprop.0]
File "", line 0, in tvm::meta_schedule::Database::JSONDatabase(tvm::ffi::String, tvm::ffi::String, bool, tvm::ffi::String)
File "", line 0, in tvm::meta_schedule::JSONFileReadLines(tvm::ffi::String const&, int, bool)
File "", line 0, in tvm::runtime::detail::LogFatal::~LogFatal() [clone .constprop.0]
File "", line 0, in tvm::runtime::detail::LogFatal::Entry::Finalize()
ValueError: Check failed: (os.good()) is false: Cannot create new file: ./tuning_logs_alexnet/database_workload.json

[gemini-code-assist]

Summary of Changes

Hello @thaisacs, I'm Gemini Code Assist¹! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed!

This pull request resolves a critical ValueError that previously occurred when initializing a JSONDatabase for model tuning if the specified database directory did not already exist. By proactively creating the necessary directory, the change ensures robust operation of the tuning process, particularly in scenarios involving zero trials, thereby improving the overall stability and user experience.

Highlights

• Database Directory Creation: Implemented a fix to ensure the database directory exists before JSONDatabase attempts to write files, preventing a ValueError when tuning models with zero trials.

Using Gemini Code Assist

The full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips.

Invoking Gemini

You can request assistance from Gemini at any point by creating a comment using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands on the current page.

Feature	Command	Description
Code Review	/gemini review	Performs a code review for the current pull request in its current state.
Pull Request Summary	/gemini summary	Provides a summary of the current pull request in its current state.
Comment	@gemini-code-assist	Responds in comments when explicitly tagged, both in pull request comments and review comments.
Help	/gemini help	Displays a list of available commands.

Customization

To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a .gemini/ folder in the base of the repository. Detailed instructions can be found here.

Limitations & Feedback

Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments. If you're interested in giving your feedback about your experience with Gemini Code Assist for Github and other Google products, sign up here.

You can also get AI-powered code generation, chat, as well as code reviews directly in the IDE at no cost with the Gemini Code Assist IDE Extension.

Footnotes

1. Review the Privacy Notices, Generative AI Prohibited Use Policy, Terms of Service, and learn how to configure Gemini Code Assist in GitHub here. Gemini can make mistakes, so double check it and use code with caution. ↩

[gemini-code-assist]

Code Review

This pull request fixes a crash that occurs when the database directory for tuning does not exist. The fix correctly introduces logic to create the directory before attempting to write database files. My review includes two suggestions to improve the implementation: one to use create_directories for better robustness in handling nested directory paths, and another to relocate the <filesystem> include to the source file where it is directly used, which is a better practice for dependency management.

[gemini-code-assist]

To improve code organization and dependency management, it's better to include <filesystem> directly in src/relax/transform/meta_schedule.cc where it's used, rather than in this public header file. This avoids unnecessarily including it for all consumers of database.h and makes dependencies clearer.

[gemini-code-assist]

This is a good fix for the reported issue. For better robustness, consider using std::filesystem::create_directories instead of create_directory. create_directories can create parent directories if they don't exist (e.g., if work_dir is a nested path like a/b/c) and it doesn't produce an error if the directory already exists.

Suggested change

	std::filesystem::create_directory(work_dir.value().c_str());
	std::filesystem::create_directories(work_dir.value().c_str());

[tlopex]

LGTM. Thanks!