diff --git a/poetry.lock b/poetry.lock index 33615a8fef..91246c0a29 100644 --- a/poetry.lock +++ b/poetry.lock @@ -9721,15 +9721,13 @@ typing_extensions = {version = ">=4.5.0", markers = "python_version >= \"3.7\""} [[package]] name = "swebench" -version = "3.0.17" +version = "4.0.3" description = "The official SWE-bench package - a benchmark for evaluating LMs on software engineering" optional = false -python-versions = ">=3.8" +python-versions = ">=3.10" groups = ["evaluation"] -files = [ - {file = "swebench-3.0.17-py3-none-any.whl", hash = "sha256:5a2f6c6d6df164d81474e6901275e73802dc5817905c82902438928544c593b3"}, - {file = "swebench-3.0.17.tar.gz", hash = "sha256:4b86d02d7bdeb2113b3c1f7759ea2992d905f0e4053be39a6d6f4573d71c9bde"}, -] +files = [] +develop = false [package.dependencies] beautifulsoup4 = "*" @@ -9749,10 +9747,16 @@ unidiff = "*" [package.extras] datasets = ["anthropic", "jedi", "openai", "protobuf", "sentencepiece", "tiktoken", "transformers"] -docs = ["griffe-pydantic", "mike", "mkdocs", "mkdocs-glightbox", "mkdocs-include-markdown-plugin", "mkdocs-material", "mkdocstrings", "mkdocstrings-python", "pymdown-extensions"] +docs = ["mike", "mkdocs", "mkdocs-glightbox", "mkdocs-include-markdown-plugin", "mkdocs-material", "mkdocstrings", "mkdocstrings-python", "pymdown-extensions"] inference = ["anthropic", "flash_attn", "jedi", "openai", "peft", "protobuf", "requests", "sentencepiece", "tiktoken", "torch", "transformers", "triton"] test = ["pytest", "pytest-cov"] +[package.source] +type = "git" +url = "https://github.com/ryanhoangt/SWE-bench.git" +reference = "fix-modal-patch-eval" +resolved_reference = "b8ffb7b48a9113de25155653de38ed6f9e54f68c" + [[package]] name = "swegym" version = "2.0.13" @@ -11795,4 +11799,4 @@ third-party-runtimes = ["daytona", "e2b", "modal", "runloop-api-client"] [metadata] lock-version = "2.1" python-versions = "^3.12,<3.14" -content-hash = "653c4cda22ec5ff95420d305386c53ba440714fe7c59a9f7f240fdf86b698031" +content-hash = "de8f45fdd525059f1c021767c7f24ef4eaf9eb00b57772c7017a86ae534c040d" diff --git a/pyproject.toml b/pyproject.toml index 6418d02988..98aeab2cc5 100644 --- a/pyproject.toml +++ b/pyproject.toml @@ -140,7 +140,6 @@ streamlit = "*" whatthepatch = "*" retry = "*" evaluate = "*" -swebench = "^3.0.8" visualswebench = { git = "https://github.com/luolin101/Visual-SWE-bench.git" } swegym = { git = "https://github.com/SWE-Gym/SWE-Bench-Package.git" } commit0 = "*" @@ -155,9 +154,11 @@ browsergym-webarena = "0.13.3" browsergym-miniwob = "0.13.3" browsergym-visualwebarena = "0.13.3" boto3-stubs = { extras = [ "s3" ], version = "^1.37.19" } -pyarrow = "20.0.0" # transitive dependency, pinned here to avoid conflicts +# transitive dependency, pinned here to avoid conflicts +pyarrow = "20.0.0" datasets = "*" joblib = "*" +swebench = { git = "https://github.com/ryanhoangt/SWE-bench.git", rev = "fix-modal-patch-eval" } [tool.poetry.scripts] openhands = "openhands.cli.main:main"