Add files using upload-large-folder tool

ckpts/llama-3b/global_step100/bf16_zero_pp_rank_154_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b875ee81b5c8375be7e5410ba10694a6d6a02a66e9f188eb085970b012668a5a
+size 41830148

ckpts/llama-3b/global_step100/bf16_zero_pp_rank_169_mp_rank_01_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f8c1c8c86830ae13ac8451ec68bcea1534070ceac5b00547b696235339aa3af6
+size 41830148

ckpts/llama-3b/global_step100/bf16_zero_pp_rank_191_mp_rank_03_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bcb402ef8e202173495e1f2984b5128dcb865c560a032f40686c0af7b3adf760
+size 41830340

ckpts/llama-3b/global_step100/bf16_zero_pp_rank_220_mp_rank_03_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1ba7f4713f5a4fd90d981efcbceb2d421b2306259122738a1d3c6eb8aa19b359
+size 41830468

venv/lib/python3.10/site-packages/pandas/_libs/byteswap.pyi

@@ -0,0 +1,5 @@
+def read_float_with_byteswap(data: bytes, offset: int, byteswap: bool) -> float: ...
+def read_double_with_byteswap(data: bytes, offset: int, byteswap: bool) -> float: ...
+def read_uint16_with_byteswap(data: bytes, offset: int, byteswap: bool) -> int: ...
+def read_uint32_with_byteswap(data: bytes, offset: int, byteswap: bool) -> int: ...
+def read_uint64_with_byteswap(data: bytes, offset: int, byteswap: bool) -> int: ...

venv/lib/python3.10/site-packages/pandas/_libs/groupby.pyi

@@ -0,0 +1,216 @@
+from typing import Literal
+
+import numpy as np
+
+from pandas._typing import npt
+
+def group_median_float64(
+    out: np.ndarray,  # ndarray[float64_t, ndim=2]
+    counts: npt.NDArray[np.int64],
+    values: np.ndarray,  # ndarray[float64_t, ndim=2]
+    labels: npt.NDArray[np.int64],
+    min_count: int = ...,  # Py_ssize_t
+    mask: np.ndarray | None = ...,
+    result_mask: np.ndarray | None = ...,
+) -> None: ...
+def group_cumprod(
+    out: np.ndarray,  # float64_t[:, ::1]
+    values: np.ndarray,  # const float64_t[:, :]
+    labels: np.ndarray,  # const int64_t[:]
+    ngroups: int,
+    is_datetimelike: bool,
+    skipna: bool = ...,
+    mask: np.ndarray | None = ...,
+    result_mask: np.ndarray | None = ...,
+) -> None: ...
+def group_cumsum(
+    out: np.ndarray,  # int64float_t[:, ::1]
+    values: np.ndarray,  # ndarray[int64float_t, ndim=2]
+    labels: np.ndarray,  # const int64_t[:]
+    ngroups: int,
+    is_datetimelike: bool,
+    skipna: bool = ...,
+    mask: np.ndarray | None = ...,
+    result_mask: np.ndarray | None = ...,
+) -> None: ...
+def group_shift_indexer(
+    out: np.ndarray,  # int64_t[::1]
+    labels: np.ndarray,  # const int64_t[:]
+    ngroups: int,
+    periods: int,
+) -> None: ...
+def group_fillna_indexer(
+    out: np.ndarray,  # ndarray[intp_t]
+    labels: np.ndarray,  # ndarray[int64_t]
+    sorted_labels: npt.NDArray[np.intp],
+    mask: npt.NDArray[np.uint8],
+    limit: int,  # int64_t
+    dropna: bool,
+) -> None: ...
+def group_any_all(
+    out: np.ndarray,  # uint8_t[::1]
+    values: np.ndarray,  # const uint8_t[::1]
+    labels: np.ndarray,  # const int64_t[:]
+    mask: np.ndarray,  # const uint8_t[::1]
+    val_test: Literal["any", "all"],
+    skipna: bool,
+    result_mask: np.ndarray | None,
+) -> None: ...
+def group_sum(
+    out: np.ndarray,  # complexfloatingintuint_t[:, ::1]
+    counts: np.ndarray,  # int64_t[::1]
+    values: np.ndarray,  # ndarray[complexfloatingintuint_t, ndim=2]
+    labels: np.ndarray,  # const intp_t[:]
+    mask: np.ndarray | None,
+    result_mask: np.ndarray | None = ...,
+    min_count: int = ...,
+    is_datetimelike: bool = ...,
+) -> None: ...
+def group_prod(
+    out: np.ndarray,  # int64float_t[:, ::1]
+    counts: np.ndarray,  # int64_t[::1]
+    values: np.ndarray,  # ndarray[int64float_t, ndim=2]
+    labels: np.ndarray,  # const intp_t[:]
+    mask: np.ndarray | None,
+    result_mask: np.ndarray | None = ...,
+    min_count: int = ...,
+) -> None: ...
+def group_var(
+    out: np.ndarray,  # floating[:, ::1]
+    counts: np.ndarray,  # int64_t[::1]
+    values: np.ndarray,  # ndarray[floating, ndim=2]
+    labels: np.ndarray,  # const intp_t[:]
+    min_count: int = ...,  # Py_ssize_t
+    ddof: int = ...,  # int64_t
+    mask: np.ndarray | None = ...,
+    result_mask: np.ndarray | None = ...,
+    is_datetimelike: bool = ...,
+    name: str = ...,
+) -> None: ...
+def group_skew(
+    out: np.ndarray,  # float64_t[:, ::1]
+    counts: np.ndarray,  # int64_t[::1]
+    values: np.ndarray,  # ndarray[float64_T, ndim=2]
+    labels: np.ndarray,  # const intp_t[::1]
+    mask: np.ndarray | None = ...,
+    result_mask: np.ndarray | None = ...,
+    skipna: bool = ...,
+) -> None: ...
+def group_mean(
+    out: np.ndarray,  # floating[:, ::1]
+    counts: np.ndarray,  # int64_t[::1]
+    values: np.ndarray,  # ndarray[floating, ndim=2]
+    labels: np.ndarray,  # const intp_t[:]
+    min_count: int = ...,  # Py_ssize_t
+    is_datetimelike: bool = ...,  # bint
+    mask: np.ndarray | None = ...,
+    result_mask: np.ndarray | None = ...,
+) -> None: ...
+def group_ohlc(
+    out: np.ndarray,  # floatingintuint_t[:, ::1]
+    counts: np.ndarray,  # int64_t[::1]
+    values: np.ndarray,  # ndarray[floatingintuint_t, ndim=2]
+    labels: np.ndarray,  # const intp_t[:]
+    min_count: int = ...,
+    mask: np.ndarray | None = ...,
+    result_mask: np.ndarray | None = ...,
+) -> None: ...
+def group_quantile(
+    out: npt.NDArray[np.float64],
+    values: np.ndarray,  # ndarray[numeric, ndim=1]
+    labels: npt.NDArray[np.intp],
+    mask: npt.NDArray[np.uint8],
+    qs: npt.NDArray[np.float64],  # const
+    starts: npt.NDArray[np.int64],
+    ends: npt.NDArray[np.int64],
+    interpolation: Literal["linear", "lower", "higher", "nearest", "midpoint"],
+    result_mask: np.ndarray | None,
+    is_datetimelike: bool,
+) -> None: ...
+def group_last(
+    out: np.ndarray,  # rank_t[:, ::1]
+    counts: np.ndarray,  # int64_t[::1]
+    values: np.ndarray,  # ndarray[rank_t, ndim=2]
+    labels: np.ndarray,  # const int64_t[:]
+    mask: npt.NDArray[np.bool_] | None,
+    result_mask: npt.NDArray[np.bool_] | None = ...,
+    min_count: int = ...,  # Py_ssize_t
+    is_datetimelike: bool = ...,
+    skipna: bool = ...,
+) -> None: ...
+def group_nth(
+    out: np.ndarray,  # rank_t[:, ::1]
+    counts: np.ndarray,  # int64_t[::1]
+    values: np.ndarray,  # ndarray[rank_t, ndim=2]
+    labels: np.ndarray,  # const int64_t[:]
+    mask: npt.NDArray[np.bool_] | None,
+    result_mask: npt.NDArray[np.bool_] | None = ...,
+    min_count: int = ...,  # int64_t
+    rank: int = ...,  # int64_t
+    is_datetimelike: bool = ...,
+    skipna: bool = ...,
+) -> None: ...
+def group_rank(
+    out: np.ndarray,  # float64_t[:, ::1]
+    values: np.ndarray,  # ndarray[rank_t, ndim=2]
+    labels: np.ndarray,  # const int64_t[:]
+    ngroups: int,
+    is_datetimelike: bool,
+    ties_method: Literal["average", "min", "max", "first", "dense"] = ...,
+    ascending: bool = ...,
+    pct: bool = ...,
+    na_option: Literal["keep", "top", "bottom"] = ...,
+    mask: npt.NDArray[np.bool_] | None = ...,
+) -> None: ...
+def group_max(
+    out: np.ndarray,  # groupby_t[:, ::1]
+    counts: np.ndarray,  # int64_t[::1]
+    values: np.ndarray,  # ndarray[groupby_t, ndim=2]
+    labels: np.ndarray,  # const int64_t[:]
+    min_count: int = ...,
+    is_datetimelike: bool = ...,
+    mask: np.ndarray | None = ...,
+    result_mask: np.ndarray | None = ...,
+) -> None: ...
+def group_min(
+    out: np.ndarray,  # groupby_t[:, ::1]
+    counts: np.ndarray,  # int64_t[::1]
+    values: np.ndarray,  # ndarray[groupby_t, ndim=2]
+    labels: np.ndarray,  # const int64_t[:]
+    min_count: int = ...,
+    is_datetimelike: bool = ...,
+    mask: np.ndarray | None = ...,
+    result_mask: np.ndarray | None = ...,
+) -> None: ...
+def group_idxmin_idxmax(
+    out: npt.NDArray[np.intp],
+    counts: npt.NDArray[np.int64],
+    values: np.ndarray,  # ndarray[groupby_t, ndim=2]
+    labels: npt.NDArray[np.intp],
+    min_count: int = ...,
+    is_datetimelike: bool = ...,
+    mask: np.ndarray | None = ...,
+    name: str = ...,
+    skipna: bool = ...,
+    result_mask: np.ndarray | None = ...,
+) -> None: ...
+def group_cummin(
+    out: np.ndarray,  # groupby_t[:, ::1]
+    values: np.ndarray,  # ndarray[groupby_t, ndim=2]
+    labels: np.ndarray,  # const int64_t[:]
+    ngroups: int,
+    is_datetimelike: bool,
+    mask: np.ndarray | None = ...,
+    result_mask: np.ndarray | None = ...,
+    skipna: bool = ...,
+) -> None: ...
+def group_cummax(
+    out: np.ndarray,  # groupby_t[:, ::1]
+    values: np.ndarray,  # ndarray[groupby_t, ndim=2]
+    labels: np.ndarray,  # const int64_t[:]
+    ngroups: int,
+    is_datetimelike: bool,
+    mask: np.ndarray | None = ...,
+    result_mask: np.ndarray | None = ...,
+    skipna: bool = ...,
+) -> None: ...

venv/lib/python3.10/site-packages/pandas/_libs/hashing.pyi

@@ -0,0 +1,9 @@
+import numpy as np
+
+from pandas._typing import npt
+
+def hash_object_array(
+    arr: npt.NDArray[np.object_],
+    key: str,
+    encoding: str = ...,
+) -> npt.NDArray[np.uint64]: ...

venv/lib/python3.10/site-packages/pandas/_libs/index.pyi

@@ -0,0 +1,100 @@
+import numpy as np
+
+from pandas._typing import npt
+
+from pandas import MultiIndex
+from pandas.core.arrays import ExtensionArray
+
+multiindex_nulls_shift: int
+
+class IndexEngine:
+    over_size_threshold: bool
+    def __init__(self, values: np.ndarray) -> None: ...
+    def __contains__(self, val: object) -> bool: ...
+
+    # -> int | slice | np.ndarray[bool]
+    def get_loc(self, val: object) -> int | slice | np.ndarray: ...
+    def sizeof(self, deep: bool = ...) -> int: ...
+    def __sizeof__(self) -> int: ...
+    @property
+    def is_unique(self) -> bool: ...
+    @property
+    def is_monotonic_increasing(self) -> bool: ...
+    @property
+    def is_monotonic_decreasing(self) -> bool: ...
+    @property
+    def is_mapping_populated(self) -> bool: ...
+    def clear_mapping(self): ...
+    def get_indexer(self, values: np.ndarray) -> npt.NDArray[np.intp]: ...
+    def get_indexer_non_unique(
+        self,
+        targets: np.ndarray,
+    ) -> tuple[npt.NDArray[np.intp], npt.NDArray[np.intp]]: ...
+
+class MaskedIndexEngine(IndexEngine):
+    def __init__(self, values: object) -> None: ...
+    def get_indexer_non_unique(
+        self, targets: object
+    ) -> tuple[npt.NDArray[np.intp], npt.NDArray[np.intp]]: ...
+
+class Float64Engine(IndexEngine): ...
+class Float32Engine(IndexEngine): ...
+class Complex128Engine(IndexEngine): ...
+class Complex64Engine(IndexEngine): ...
+class Int64Engine(IndexEngine): ...
+class Int32Engine(IndexEngine): ...
+class Int16Engine(IndexEngine): ...
+class Int8Engine(IndexEngine): ...
+class UInt64Engine(IndexEngine): ...
+class UInt32Engine(IndexEngine): ...
+class UInt16Engine(IndexEngine): ...
+class UInt8Engine(IndexEngine): ...
+class ObjectEngine(IndexEngine): ...
+class DatetimeEngine(Int64Engine): ...
+class TimedeltaEngine(DatetimeEngine): ...
+class PeriodEngine(Int64Engine): ...
+class BoolEngine(UInt8Engine): ...
+class MaskedFloat64Engine(MaskedIndexEngine): ...
+class MaskedFloat32Engine(MaskedIndexEngine): ...
+class MaskedComplex128Engine(MaskedIndexEngine): ...
+class MaskedComplex64Engine(MaskedIndexEngine): ...
+class MaskedInt64Engine(MaskedIndexEngine): ...
+class MaskedInt32Engine(MaskedIndexEngine): ...
+class MaskedInt16Engine(MaskedIndexEngine): ...
+class MaskedInt8Engine(MaskedIndexEngine): ...
+class MaskedUInt64Engine(MaskedIndexEngine): ...
+class MaskedUInt32Engine(MaskedIndexEngine): ...
+class MaskedUInt16Engine(MaskedIndexEngine): ...
+class MaskedUInt8Engine(MaskedIndexEngine): ...
+class MaskedBoolEngine(MaskedUInt8Engine): ...
+
+class BaseMultiIndexCodesEngine:
+    levels: list[np.ndarray]
+    offsets: np.ndarray  # ndarray[uint64_t, ndim=1]
+
+    def __init__(
+        self,
+        levels: list[np.ndarray],  # all entries hashable
+        labels: list[np.ndarray],  # all entries integer-dtyped
+        offsets: np.ndarray,  # np.ndarray[np.uint64, ndim=1]
+    ) -> None: ...
+    def get_indexer(self, target: npt.NDArray[np.object_]) -> npt.NDArray[np.intp]: ...
+    def _extract_level_codes(self, target: MultiIndex) -> np.ndarray: ...
+
+class ExtensionEngine:
+    def __init__(self, values: ExtensionArray) -> None: ...
+    def __contains__(self, val: object) -> bool: ...
+    def get_loc(self, val: object) -> int | slice | np.ndarray: ...
+    def get_indexer(self, values: np.ndarray) -> npt.NDArray[np.intp]: ...
+    def get_indexer_non_unique(
+        self,
+        targets: np.ndarray,
+    ) -> tuple[npt.NDArray[np.intp], npt.NDArray[np.intp]]: ...
+    @property
+    def is_unique(self) -> bool: ...
+    @property
+    def is_monotonic_increasing(self) -> bool: ...
+    @property
+    def is_monotonic_decreasing(self) -> bool: ...
+    def sizeof(self, deep: bool = ...) -> int: ...
+    def clear_mapping(self): ...

venv/lib/python3.10/site-packages/pandas/_libs/internals.pyi

@@ -0,0 +1,94 @@
+from typing import (
+    Iterator,
+    Sequence,
+    final,
+    overload,
+)
+import weakref
+
+import numpy as np
+
+from pandas._typing import (
+    ArrayLike,
+    Self,
+    npt,
+)
+
+from pandas import Index
+from pandas.core.internals.blocks import Block as B
+
+def slice_len(slc: slice, objlen: int = ...) -> int: ...
+def get_concat_blkno_indexers(
+    blknos_list: list[npt.NDArray[np.intp]],
+) -> list[tuple[npt.NDArray[np.intp], BlockPlacement]]: ...
+def get_blkno_indexers(
+    blknos: np.ndarray,  # int64_t[:]
+    group: bool = ...,
+) -> list[tuple[int, slice | np.ndarray]]: ...
+def get_blkno_placements(
+    blknos: np.ndarray,
+    group: bool = ...,
+) -> Iterator[tuple[int, BlockPlacement]]: ...
+def update_blklocs_and_blknos(
+    blklocs: npt.NDArray[np.intp],
+    blknos: npt.NDArray[np.intp],
+    loc: int,
+    nblocks: int,
+) -> tuple[npt.NDArray[np.intp], npt.NDArray[np.intp]]: ...
+@final
+class BlockPlacement:
+    def __init__(self, val: int | slice | np.ndarray) -> None: ...
+    @property
+    def indexer(self) -> np.ndarray | slice: ...
+    @property
+    def as_array(self) -> np.ndarray: ...
+    @property
+    def as_slice(self) -> slice: ...
+    @property
+    def is_slice_like(self) -> bool: ...
+    @overload
+    def __getitem__(
+        self, loc: slice | Sequence[int] | npt.NDArray[np.intp]
+    ) -> BlockPlacement: ...
+    @overload
+    def __getitem__(self, loc: int) -> int: ...
+    def __iter__(self) -> Iterator[int]: ...
+    def __len__(self) -> int: ...
+    def delete(self, loc) -> BlockPlacement: ...
+    def add(self, other) -> BlockPlacement: ...
+    def append(self, others: list[BlockPlacement]) -> BlockPlacement: ...
+    def tile_for_unstack(self, factor: int) -> npt.NDArray[np.intp]: ...
+
+class Block:
+    _mgr_locs: BlockPlacement
+    ndim: int
+    values: ArrayLike
+    refs: BlockValuesRefs
+    def __init__(
+        self,
+        values: ArrayLike,
+        placement: BlockPlacement,
+        ndim: int,
+        refs: BlockValuesRefs | None = ...,
+    ) -> None: ...
+    def slice_block_rows(self, slicer: slice) -> Self: ...
+
+class BlockManager:
+    blocks: tuple[B, ...]
+    axes: list[Index]
+    _known_consolidated: bool
+    _is_consolidated: bool
+    _blknos: np.ndarray
+    _blklocs: np.ndarray
+    def __init__(
+        self, blocks: tuple[B, ...], axes: list[Index], verify_integrity=...
+    ) -> None: ...
+    def get_slice(self, slobj: slice, axis: int = ...) -> Self: ...
+    def _rebuild_blknos_and_blklocs(self) -> None: ...
+
+class BlockValuesRefs:
+    referenced_blocks: list[weakref.ref]
+    def __init__(self, blk: Block | None = ...) -> None: ...
+    def add_reference(self, blk: Block) -> None: ...
+    def add_index_reference(self, index: Index) -> None: ...
+    def has_reference(self) -> bool: ...

venv/lib/python3.10/site-packages/pandas/_libs/sparse.pyi

@@ -0,0 +1,51 @@
+from typing import Sequence
+
+import numpy as np
+
+from pandas._typing import (
+    Self,
+    npt,
+)
+
+class SparseIndex:
+    length: int
+    npoints: int
+    def __init__(self) -> None: ...
+    @property
+    def ngaps(self) -> int: ...
+    @property
+    def nbytes(self) -> int: ...
+    @property
+    def indices(self) -> npt.NDArray[np.int32]: ...
+    def equals(self, other) -> bool: ...
+    def lookup(self, index: int) -> np.int32: ...
+    def lookup_array(self, indexer: npt.NDArray[np.int32]) -> npt.NDArray[np.int32]: ...
+    def to_int_index(self) -> IntIndex: ...
+    def to_block_index(self) -> BlockIndex: ...
+    def intersect(self, y_: SparseIndex) -> Self: ...
+    def make_union(self, y_: SparseIndex) -> Self: ...
+
+class IntIndex(SparseIndex):
+    indices: npt.NDArray[np.int32]
+    def __init__(
+        self, length: int, indices: Sequence[int], check_integrity: bool = ...
+    ) -> None: ...
+
+class BlockIndex(SparseIndex):
+    nblocks: int
+    blocs: np.ndarray
+    blengths: np.ndarray
+    def __init__(
+        self, length: int, blocs: np.ndarray, blengths: np.ndarray
+    ) -> None: ...
+
+    # Override to have correct parameters
+    def intersect(self, other: SparseIndex) -> Self: ...
+    def make_union(self, y: SparseIndex) -> Self: ...
+
+def make_mask_object_ndarray(
+    arr: npt.NDArray[np.object_], fill_value
+) -> npt.NDArray[np.bool_]: ...
+def get_blocks(
+    indices: npt.NDArray[np.int32],
+) -> tuple[npt.NDArray[np.int32], npt.NDArray[np.int32]]: ...

venv/lib/python3.10/site-packages/pandas/_libs/tslib.pyi

@@ -0,0 +1,37 @@
+from datetime import tzinfo
+
+import numpy as np
+
+from pandas._typing import npt
+
+def format_array_from_datetime(
+    values: npt.NDArray[np.int64],
+    tz: tzinfo | None = ...,
+    format: str | None = ...,
+    na_rep: str | float = ...,
+    reso: int = ...,  # NPY_DATETIMEUNIT
+) -> npt.NDArray[np.object_]: ...
+def array_with_unit_to_datetime(
+    values: npt.NDArray[np.object_],
+    unit: str,
+    errors: str = ...,
+) -> tuple[np.ndarray, tzinfo | None]: ...
+def first_non_null(values: np.ndarray) -> int: ...
+def array_to_datetime(
+    values: npt.NDArray[np.object_],
+    errors: str = ...,
+    dayfirst: bool = ...,
+    yearfirst: bool = ...,
+    utc: bool = ...,
+    creso: int = ...,
+) -> tuple[np.ndarray, tzinfo | None]: ...
+
+# returned ndarray may be object dtype or datetime64[ns]
+
+def array_to_datetime_with_tz(
+    values: npt.NDArray[np.object_],
+    tz: tzinfo,
+    dayfirst: bool,
+    yearfirst: bool,
+    creso: int,
+) -> npt.NDArray[np.int64]: ...

venv/lib/python3.10/site-packages/pandas/core/_numba/executor.py

@@ -0,0 +1,239 @@
+from __future__ import annotations
+
+import functools
+from typing import (
+    TYPE_CHECKING,
+    Any,
+    Callable,
+)
+
+if TYPE_CHECKING:
+    from pandas._typing import Scalar
+
+import numpy as np
+
+from pandas.compat._optional import import_optional_dependency
+
+
+@functools.cache
+def generate_apply_looper(func, nopython=True, nogil=True, parallel=False):
+    if TYPE_CHECKING:
+        import numba
+    else:
+        numba = import_optional_dependency("numba")
+    nb_compat_func = numba.extending.register_jitable(func)
+
+    @numba.jit(nopython=nopython, nogil=nogil, parallel=parallel)
+    def nb_looper(values, axis):
+        # Operate on the first row/col in order to get
+        # the output shape
+        if axis == 0:
+            first_elem = values[:, 0]
+            dim0 = values.shape[1]
+        else:
+            first_elem = values[0]
+            dim0 = values.shape[0]
+        res0 = nb_compat_func(first_elem)
+        # Use np.asarray to get shape for
+        # https://github.com/numba/numba/issues/4202#issuecomment-1185981507
+        buf_shape = (dim0,) + np.atleast_1d(np.asarray(res0)).shape
+        if axis == 0:
+            buf_shape = buf_shape[::-1]
+        buff = np.empty(buf_shape)
+
+        if axis == 1:
+            buff[0] = res0
+            for i in numba.prange(1, values.shape[0]):
+                buff[i] = nb_compat_func(values[i])
+        else:
+            buff[:, 0] = res0
+            for j in numba.prange(1, values.shape[1]):
+                buff[:, j] = nb_compat_func(values[:, j])
+        return buff
+
+    return nb_looper
+
+
+@functools.cache
+def make_looper(func, result_dtype, is_grouped_kernel, nopython, nogil, parallel):
+    if TYPE_CHECKING:
+        import numba
+    else:
+        numba = import_optional_dependency("numba")
+
+    if is_grouped_kernel:
+
+        @numba.jit(nopython=nopython, nogil=nogil, parallel=parallel)
+        def column_looper(
+            values: np.ndarray,
+            labels: np.ndarray,
+            ngroups: int,
+            min_periods: int,
+            *args,
+        ):
+            result = np.empty((values.shape[0], ngroups), dtype=result_dtype)
+            na_positions = {}
+            for i in numba.prange(values.shape[0]):
+                output, na_pos = func(
+                    values[i], result_dtype, labels, ngroups, min_periods, *args
+                )
+                result[i] = output
+                if len(na_pos) > 0:
+                    na_positions[i] = np.array(na_pos)
+            return result, na_positions
+
+    else:
+
+        @numba.jit(nopython=nopython, nogil=nogil, parallel=parallel)
+        def column_looper(
+            values: np.ndarray,
+            start: np.ndarray,
+            end: np.ndarray,
+            min_periods: int,
+            *args,
+        ):
+            result = np.empty((values.shape[0], len(start)), dtype=result_dtype)
+            na_positions = {}
+            for i in numba.prange(values.shape[0]):
+                output, na_pos = func(
+                    values[i], result_dtype, start, end, min_periods, *args
+                )
+                result[i] = output
+                if len(na_pos) > 0:
+                    na_positions[i] = np.array(na_pos)
+            return result, na_positions
+
+    return column_looper
+
+
+default_dtype_mapping: dict[np.dtype, Any] = {
+    np.dtype("int8"): np.int64,
+    np.dtype("int16"): np.int64,
+    np.dtype("int32"): np.int64,
+    np.dtype("int64"): np.int64,
+    np.dtype("uint8"): np.uint64,
+    np.dtype("uint16"): np.uint64,
+    np.dtype("uint32"): np.uint64,
+    np.dtype("uint64"): np.uint64,
+    np.dtype("float32"): np.float64,
+    np.dtype("float64"): np.float64,
+    np.dtype("complex64"): np.complex128,
+    np.dtype("complex128"): np.complex128,
+}
+
+
+# TODO: Preserve complex dtypes
+
+float_dtype_mapping: dict[np.dtype, Any] = {
+    np.dtype("int8"): np.float64,
+    np.dtype("int16"): np.float64,
+    np.dtype("int32"): np.float64,
+    np.dtype("int64"): np.float64,
+    np.dtype("uint8"): np.float64,
+    np.dtype("uint16"): np.float64,
+    np.dtype("uint32"): np.float64,
+    np.dtype("uint64"): np.float64,
+    np.dtype("float32"): np.float64,
+    np.dtype("float64"): np.float64,
+    np.dtype("complex64"): np.float64,
+    np.dtype("complex128"): np.float64,
+}
+
+identity_dtype_mapping: dict[np.dtype, Any] = {
+    np.dtype("int8"): np.int8,
+    np.dtype("int16"): np.int16,
+    np.dtype("int32"): np.int32,
+    np.dtype("int64"): np.int64,
+    np.dtype("uint8"): np.uint8,
+    np.dtype("uint16"): np.uint16,
+    np.dtype("uint32"): np.uint32,
+    np.dtype("uint64"): np.uint64,
+    np.dtype("float32"): np.float32,
+    np.dtype("float64"): np.float64,
+    np.dtype("complex64"): np.complex64,
+    np.dtype("complex128"): np.complex128,
+}
+
+
+def generate_shared_aggregator(
+    func: Callable[..., Scalar],
+    dtype_mapping: dict[np.dtype, np.dtype],
+    is_grouped_kernel: bool,
+    nopython: bool,
+    nogil: bool,
+    parallel: bool,
+):
+    """
+    Generate a Numba function that loops over the columns 2D object and applies
+    a 1D numba kernel over each column.
+
+    Parameters
+    ----------
+    func : function
+        aggregation function to be applied to each column
+    dtype_mapping: dict or None
+        If not None, maps a dtype to a result dtype.
+        Otherwise, will fall back to default mapping.
+    is_grouped_kernel: bool, default False
+        Whether func operates using the group labels (True)
+        or using starts/ends arrays
+
+        If true, you also need to pass the number of groups to this function
+    nopython : bool
+        nopython to be passed into numba.jit
+    nogil : bool
+        nogil to be passed into numba.jit
+    parallel : bool
+        parallel to be passed into numba.jit
+
+    Returns
+    -------
+    Numba function
+    """
+
+    # A wrapper around the looper function,
+    # to dispatch based on dtype since numba is unable to do that in nopython mode
+
+    # It also post-processes the values by inserting nans where number of observations
+    # is less than min_periods
+    # Cannot do this in numba nopython mode
+    # (you'll run into type-unification error when you cast int -> float)
+    def looper_wrapper(
+        values,
+        start=None,
+        end=None,
+        labels=None,
+        ngroups=None,
+        min_periods: int = 0,
+        **kwargs,
+    ):
+        result_dtype = dtype_mapping[values.dtype]
+        column_looper = make_looper(
+            func, result_dtype, is_grouped_kernel, nopython, nogil, parallel
+        )
+        # Need to unpack kwargs since numba only supports *args
+        if is_grouped_kernel:
+            result, na_positions = column_looper(
+                values, labels, ngroups, min_periods, *kwargs.values()
+            )
+        else:
+            result, na_positions = column_looper(
+                values, start, end, min_periods, *kwargs.values()
+            )
+        if result.dtype.kind == "i":
+            # Look if na_positions is not empty
+            # If so, convert the whole block
+            # This is OK since int dtype cannot hold nan,
+            # so if min_periods not satisfied for 1 col, it is not satisfied for
+            # all columns at that index
+            for na_pos in na_positions.values():
+                if len(na_pos) > 0:
+                    result = result.astype("float64")
+                    break
+        # TODO: Optimize this
+        for i, na_pos in na_positions.items():
+            if len(na_pos) > 0:
+                result[i, na_pos] = np.nan
+        return result
+
+    return looper_wrapper

venv/lib/python3.10/site-packages/pandas/core/_numba/extensions.py

@@ -0,0 +1,584 @@
+# Disable type checking for this module since numba's internals
+# are not typed, and we use numba's internals via its extension API
+# mypy: ignore-errors
+"""
+Utility classes/functions to let numba recognize
+pandas Index/Series/DataFrame
+
+Mostly vendored from https://github.com/numba/numba/blob/main/numba/tests/pdlike_usecase.py
+"""
+
+from __future__ import annotations
+
+from contextlib import contextmanager
+import operator
+
+import numba
+from numba import types
+from numba.core import cgutils
+from numba.core.datamodel import models
+from numba.core.extending import (
+    NativeValue,
+    box,
+    lower_builtin,
+    make_attribute_wrapper,
+    overload,
+    overload_attribute,
+    overload_method,
+    register_model,
+    type_callable,
+    typeof_impl,
+    unbox,
+)
+from numba.core.imputils import impl_ret_borrowed
+import numpy as np
+
+from pandas._libs import lib
+
+from pandas.core.indexes.base import Index
+from pandas.core.indexing import _iLocIndexer
+from pandas.core.internals import SingleBlockManager
+from pandas.core.series import Series
+
+
+# Helper function to hack around fact that Index casts numpy string dtype to object
+#
+# Idea is to set an attribute on a Index called _numba_data
+# that is the original data, or the object data casted to numpy string dtype,
+# with a context manager that is unset afterwards
+@contextmanager
+def set_numba_data(index: Index):
+    numba_data = index._data
+    if numba_data.dtype == object:
+        if not lib.is_string_array(numba_data):
+            raise ValueError(
+                "The numba engine only supports using string or numeric column names"
+            )
+        numba_data = numba_data.astype("U")
+    try:
+        index._numba_data = numba_data
+        yield index
+    finally:
+        del index._numba_data
+
+
+# TODO: Range index support
+# (this currently lowers OK, but does not round-trip)
+class IndexType(types.Type):
+    """
+    The type class for Index objects.
+    """
+
+    def __init__(self, dtype, layout, pyclass: any) -> None:
+        self.pyclass = pyclass
+        name = f"index({dtype}, {layout})"
+        self.dtype = dtype
+        self.layout = layout
+        super().__init__(name)
+
+    @property
+    def key(self):
+        return self.pyclass, self.dtype, self.layout
+
+    @property
+    def as_array(self):
+        return types.Array(self.dtype, 1, self.layout)
+
+    def copy(self, dtype=None, ndim: int = 1, layout=None):
+        assert ndim == 1
+        if dtype is None:
+            dtype = self.dtype
+        layout = layout or self.layout
+        return type(self)(dtype, layout, self.pyclass)
+
+
+class SeriesType(types.Type):
+    """
+    The type class for Series objects.
+    """
+
+    def __init__(self, dtype, index, namety) -> None:
+        assert isinstance(index, IndexType)
+        self.dtype = dtype
+        self.index = index
+        self.values = types.Array(self.dtype, 1, "C")
+        self.namety = namety
+        name = f"series({dtype}, {index}, {namety})"
+        super().__init__(name)
+
+    @property
+    def key(self):
+        return self.dtype, self.index, self.namety
+
+    @property
+    def as_array(self):
+        return self.values
+
+    def copy(self, dtype=None, ndim: int = 1, layout: str = "C"):
+        assert ndim == 1
+        assert layout == "C"
+        if dtype is None:
+            dtype = self.dtype
+        return type(self)(dtype, self.index, self.namety)
+
+
+@typeof_impl.register(Index)
+def typeof_index(val, c):
+    """
+    This will assume that only strings are in object dtype
+    index.
+    (you should check this before this gets lowered down to numba)
+    """
+    # arrty = typeof_impl(val._data, c)
+    arrty = typeof_impl(val._numba_data, c)
+    assert arrty.ndim == 1
+    return IndexType(arrty.dtype, arrty.layout, type(val))
+
+
+@typeof_impl.register(Series)
+def typeof_series(val, c):
+    index = typeof_impl(val.index, c)
+    arrty = typeof_impl(val.values, c)
+    namety = typeof_impl(val.name, c)
+    assert arrty.ndim == 1
+    assert arrty.layout == "C"
+    return SeriesType(arrty.dtype, index, namety)
+
+
+@type_callable(Series)
+def type_series_constructor(context):
+    def typer(data, index, name=None):
+        if isinstance(index, IndexType) and isinstance(data, types.Array):
+            assert data.ndim == 1
+            if name is None:
+                name = types.intp
+            return SeriesType(data.dtype, index, name)
+
+    return typer
+
+
+@type_callable(Index)
+def type_index_constructor(context):
+    def typer(data, hashmap=None):
+        if isinstance(data, types.Array):
+            assert data.layout == "C"
+            assert data.ndim == 1
+            assert hashmap is None or isinstance(hashmap, types.DictType)
+            return IndexType(data.dtype, layout=data.layout, pyclass=Index)
+
+    return typer
+
+
+# Backend extensions for Index and Series and Frame
+@register_model(IndexType)
+class IndexModel(models.StructModel):
+    def __init__(self, dmm, fe_type) -> None:
+        # We don't want the numpy string scalar type in our hashmap
+        members = [
+            ("data", fe_type.as_array),
+            # This is an attempt to emulate our hashtable code with a numba
+            # typed dict
+            # It maps from values in the index to their integer positions in the array
+            ("hashmap", types.DictType(fe_type.dtype, types.intp)),
+            # Pointer to the Index object this was created from, or that it
+            # boxes to
+            # https://numba.discourse.group/t/qst-how-to-cache-the-boxing-of-an-object/2128/2?u=lithomas1
+            ("parent", types.pyobject),
+        ]
+        models.StructModel.__init__(self, dmm, fe_type, members)
+
+
+@register_model(SeriesType)
+class SeriesModel(models.StructModel):
+    def __init__(self, dmm, fe_type) -> None:
+        members = [
+            ("index", fe_type.index),
+            ("values", fe_type.as_array),
+            ("name", fe_type.namety),
+        ]
+        models.StructModel.__init__(self, dmm, fe_type, members)
+
+
+make_attribute_wrapper(IndexType, "data", "_data")
+make_attribute_wrapper(IndexType, "hashmap", "hashmap")
+
+make_attribute_wrapper(SeriesType, "index", "index")
+make_attribute_wrapper(SeriesType, "values", "values")
+make_attribute_wrapper(SeriesType, "name", "name")
+
+
+@lower_builtin(Series, types.Array, IndexType)
+def pdseries_constructor(context, builder, sig, args):
+    data, index = args
+    series = cgutils.create_struct_proxy(sig.return_type)(context, builder)
+    series.index = index
+    series.values = data
+    series.name = context.get_constant(types.intp, 0)
+    return impl_ret_borrowed(context, builder, sig.return_type, series._getvalue())
+
+
+@lower_builtin(Series, types.Array, IndexType, types.intp)
+@lower_builtin(Series, types.Array, IndexType, types.float64)
+@lower_builtin(Series, types.Array, IndexType, types.unicode_type)
+def pdseries_constructor_with_name(context, builder, sig, args):
+    data, index, name = args
+    series = cgutils.create_struct_proxy(sig.return_type)(context, builder)
+    series.index = index
+    series.values = data
+    series.name = name
+    return impl_ret_borrowed(context, builder, sig.return_type, series._getvalue())
+
+
+@lower_builtin(Index, types.Array, types.DictType, types.pyobject)
+def index_constructor_2arg(context, builder, sig, args):
+    (data, hashmap, parent) = args
+    index = cgutils.create_struct_proxy(sig.return_type)(context, builder)
+
+    index.data = data
+    index.hashmap = hashmap
+    index.parent = parent
+    return impl_ret_borrowed(context, builder, sig.return_type, index._getvalue())
+
+
+@lower_builtin(Index, types.Array, types.DictType)
+def index_constructor_2arg_parent(context, builder, sig, args):
+    # Basically same as index_constructor_1arg, but also lets you specify the
+    # parent object
+    (data, hashmap) = args
+    index = cgutils.create_struct_proxy(sig.return_type)(context, builder)
+
+    index.data = data
+    index.hashmap = hashmap
+    return impl_ret_borrowed(context, builder, sig.return_type, index._getvalue())
+
+
+@lower_builtin(Index, types.Array)
+def index_constructor_1arg(context, builder, sig, args):
+    from numba.typed import Dict
+
+    key_type = sig.return_type.dtype
+    value_type = types.intp
+
+    def index_impl(data):
+        return Index(data, Dict.empty(key_type, value_type))
+
+    return context.compile_internal(builder, index_impl, sig, args)
+
+
+# Helper to convert the unicodecharseq (numpy string scalar) into a unicode_type
+# (regular string)
+def maybe_cast_str(x):
+    # Dummy function that numba can overload
+    pass
+
+
+@overload(maybe_cast_str)
+def maybe_cast_str_impl(x):
+    """Converts numba UnicodeCharSeq (numpy string scalar) -> unicode type (string).
+    Is a no-op for other types."""
+    if isinstance(x, types.UnicodeCharSeq):
+        return lambda x: str(x)
+    else:
+        return lambda x: x
+
+
+@unbox(IndexType)
+def unbox_index(typ, obj, c):
+    """
+    Convert a Index object to a native structure.
+
+    Note: Object dtype is not allowed here
+    """
+    data_obj = c.pyapi.object_getattr_string(obj, "_numba_data")
+    index = cgutils.create_struct_proxy(typ)(c.context, c.builder)
+    # If we see an object array, assume its been validated as only containing strings
+    # We still need to do the conversion though
+    index.data = c.unbox(typ.as_array, data_obj).value
+    typed_dict_obj = c.pyapi.unserialize(c.pyapi.serialize_object(numba.typed.Dict))
+    # Create an empty typed dict in numba for the hashmap for indexing
+    # equiv of numba.typed.Dict.empty(typ.dtype, types.intp)
+    arr_type_obj = c.pyapi.unserialize(c.pyapi.serialize_object(typ.dtype))
+    intp_type_obj = c.pyapi.unserialize(c.pyapi.serialize_object(types.intp))
+    hashmap_obj = c.pyapi.call_method(
+        typed_dict_obj, "empty", (arr_type_obj, intp_type_obj)
+    )
+    index.hashmap = c.unbox(types.DictType(typ.dtype, types.intp), hashmap_obj).value
+    # Set the parent for speedy boxing.
+    index.parent = obj
+
+    # Decrefs
+    c.pyapi.decref(data_obj)
+    c.pyapi.decref(arr_type_obj)
+    c.pyapi.decref(intp_type_obj)
+    c.pyapi.decref(typed_dict_obj)
+
+    return NativeValue(index._getvalue())
+
+
+@unbox(SeriesType)
+def unbox_series(typ, obj, c):
+    """
+    Convert a Series object to a native structure.
+    """
+    index_obj = c.pyapi.object_getattr_string(obj, "index")
+    values_obj = c.pyapi.object_getattr_string(obj, "values")
+    name_obj = c.pyapi.object_getattr_string(obj, "name")
+
+    series = cgutils.create_struct_proxy(typ)(c.context, c.builder)
+    series.index = c.unbox(typ.index, index_obj).value
+    series.values = c.unbox(typ.values, values_obj).value
+    series.name = c.unbox(typ.namety, name_obj).value
+
+    # Decrefs
+    c.pyapi.decref(index_obj)
+    c.pyapi.decref(values_obj)
+    c.pyapi.decref(name_obj)
+
+    return NativeValue(series._getvalue())
+
+
+@box(IndexType)
+def box_index(typ, val, c):
+    """
+    Convert a native index structure to a Index object.
+
+    If our native index is of a numpy string dtype, we'll cast it to
+    object.
+    """
+    # First build a Numpy array object, then wrap it in a Index
+    index = cgutils.create_struct_proxy(typ)(c.context, c.builder, value=val)
+
+    res = cgutils.alloca_once_value(c.builder, index.parent)
+
+    # Does parent exist?
+    # (it means already boxed once, or Index same as original df.index or df.columns)
+    # xref https://github.com/numba/numba/blob/596e8a55334cc46854e3192766e643767bd7c934/numba/core/boxing.py#L593C17-L593C17
+    with c.builder.if_else(cgutils.is_not_null(c.builder, index.parent)) as (
+        has_parent,
+        otherwise,
+    ):
+        with has_parent:
+            c.pyapi.incref(index.parent)
+        with otherwise:
+            # TODO: preserve the original class for the index
+            # Also need preserve the name of the Index
+            # class_obj = c.pyapi.unserialize(c.pyapi.serialize_object(typ.pyclass))
+            class_obj = c.pyapi.unserialize(c.pyapi.serialize_object(Index))
+            array_obj = c.box(typ.as_array, index.data)
+            if isinstance(typ.dtype, types.UnicodeCharSeq):
+                # We converted to numpy string dtype, convert back
+                # to object since _simple_new won't do that for uss
+                object_str_obj = c.pyapi.unserialize(c.pyapi.serialize_object("object"))
+                array_obj = c.pyapi.call_method(array_obj, "astype", (object_str_obj,))
+                c.pyapi.decref(object_str_obj)
+            # this is basically Index._simple_new(array_obj, name_obj) in python
+            index_obj = c.pyapi.call_method(class_obj, "_simple_new", (array_obj,))
+            index.parent = index_obj
+            c.builder.store(index_obj, res)
+
+            # Decrefs
+            c.pyapi.decref(class_obj)
+            c.pyapi.decref(array_obj)
+    return c.builder.load(res)
+
+
+@box(SeriesType)
+def box_series(typ, val, c):
+    """
+    Convert a native series structure to a Series object.
+    """
+    series = cgutils.create_struct_proxy(typ)(c.context, c.builder, value=val)
+    series_const_obj = c.pyapi.unserialize(c.pyapi.serialize_object(Series._from_mgr))
+    mgr_const_obj = c.pyapi.unserialize(
+        c.pyapi.serialize_object(SingleBlockManager.from_array)
+    )
+    index_obj = c.box(typ.index, series.index)
+    array_obj = c.box(typ.as_array, series.values)
+    name_obj = c.box(typ.namety, series.name)
+    # This is basically equivalent of
+    # pd.Series(data=array_obj, index=index_obj)
+    # To improve perf, we will construct the Series from a manager
+    # object to avoid checks.
+    # We'll also set the name attribute manually to avoid validation
+    mgr_obj = c.pyapi.call_function_objargs(
+        mgr_const_obj,
+        (
+            array_obj,
+            index_obj,
+        ),
+    )
+    mgr_axes_obj = c.pyapi.object_getattr_string(mgr_obj, "axes")
+    # Series._constructor_from_mgr(mgr, axes)
+    series_obj = c.pyapi.call_function_objargs(
+        series_const_obj, (mgr_obj, mgr_axes_obj)
+    )
+    c.pyapi.object_setattr_string(series_obj, "_name", name_obj)
+
+    # Decrefs
+    c.pyapi.decref(series_const_obj)
+    c.pyapi.decref(mgr_axes_obj)
+    c.pyapi.decref(mgr_obj)
+    c.pyapi.decref(mgr_const_obj)
+    c.pyapi.decref(index_obj)
+    c.pyapi.decref(array_obj)
+    c.pyapi.decref(name_obj)
+
+    return series_obj
+
+
+# Add common series reductions (e.g. mean, sum),
+# and also add common binops (e.g. add, sub, mul, div)
+def generate_series_reduction(ser_reduction, ser_method):
+    @overload_method(SeriesType, ser_reduction)
+    def series_reduction(series):
+        def series_reduction_impl(series):
+            return ser_method(series.values)
+
+        return series_reduction_impl
+
+    return series_reduction
+
+
+def generate_series_binop(binop):
+    @overload(binop)
+    def series_binop(series1, value):
+        if isinstance(series1, SeriesType):
+            if isinstance(value, SeriesType):
+
+                def series_binop_impl(series1, series2):
+                    # TODO: Check index matching?
+                    return Series(
+                        binop(series1.values, series2.values),
+                        series1.index,
+                        series1.name,
+                    )
+
+                return series_binop_impl
+            else:
+
+                def series_binop_impl(series1, value):
+                    return Series(
+                        binop(series1.values, value), series1.index, series1.name
+                    )
+
+                return series_binop_impl
+
+    return series_binop
+
+
+series_reductions = [
+    ("sum", np.sum),
+    ("mean", np.mean),
+    # Disabled due to discrepancies between numba std. dev
+    # and pandas std. dev (no way to specify dof)
+    # ("std", np.std),
+    # ("var", np.var),
+    ("min", np.min),
+    ("max", np.max),
+]
+for reduction, reduction_method in series_reductions:
+    generate_series_reduction(reduction, reduction_method)
+
+series_binops = [operator.add, operator.sub, operator.mul, operator.truediv]
+
+for ser_binop in series_binops:
+    generate_series_binop(ser_binop)
+
+
+# get_loc on Index
+@overload_method(IndexType, "get_loc")
+def index_get_loc(index, item):
+    def index_get_loc_impl(index, item):
+        # Initialize the hash table if not initialized
+        if len(index.hashmap) == 0:
+            for i, val in enumerate(index._data):
+                index.hashmap[val] = i
+        return index.hashmap[item]
+
+    return index_get_loc_impl
+
+
+# Indexing for Series/Index
+@overload(operator.getitem)
+def series_indexing(series, item):
+    if isinstance(series, SeriesType):
+
+        def series_getitem(series, item):
+            loc = series.index.get_loc(item)
+            return series.iloc[loc]
+
+        return series_getitem
+
+
+@overload(operator.getitem)
+def index_indexing(index, idx):
+    if isinstance(index, IndexType):
+
+        def index_getitem(index, idx):
+            return index._data[idx]
+
+        return index_getitem
+
+
+class IlocType(types.Type):
+    def __init__(self, obj_type) -> None:
+        self.obj_type = obj_type
+        name = f"iLocIndexer({obj_type})"
+        super().__init__(name=name)
+
+    @property
+    def key(self):
+        return self.obj_type
+
+
+@typeof_impl.register(_iLocIndexer)
+def typeof_iloc(val, c):
+    objtype = typeof_impl(val.obj, c)
+    return IlocType(objtype)
+
+
+@type_callable(_iLocIndexer)
+def type_iloc_constructor(context):
+    def typer(obj):
+        if isinstance(obj, SeriesType):
+            return IlocType(obj)
+
+    return typer
+
+
+@lower_builtin(_iLocIndexer, SeriesType)
+def iloc_constructor(context, builder, sig, args):
+    (obj,) = args
+    iloc_indexer = cgutils.create_struct_proxy(sig.return_type)(context, builder)
+    iloc_indexer.obj = obj
+    return impl_ret_borrowed(
+        context, builder, sig.return_type, iloc_indexer._getvalue()
+    )
+
+
+@register_model(IlocType)
+class ILocModel(models.StructModel):
+    def __init__(self, dmm, fe_type) -> None:
+        members = [("obj", fe_type.obj_type)]
+        models.StructModel.__init__(self, dmm, fe_type, members)
+
+
+make_attribute_wrapper(IlocType, "obj", "obj")
+
+
+@overload_attribute(SeriesType, "iloc")
+def series_iloc(series):
+    def get(series):
+        return _iLocIndexer(series)
+
+    return get
+
+
+@overload(operator.getitem)
+def iloc_getitem(iloc_indexer, i):
+    if isinstance(iloc_indexer, IlocType):
+
+        def getitem_impl(iloc_indexer, i):
+            return iloc_indexer.obj.values[i]
+
+        return getitem_impl