Feature support matrix

Note

This page is generated from dev/backend_support_data.json, which is produced by dev/backend_support_sweep.py — an automated sweep that invokes every public function in saiunit.math, saiunit.linalg, saiunit.fft, and every public Quantity method across each locally-installed backend and records the outcome. Re-run the sweep and the renderer to refresh this page.

Cell legend

Glyph	Meaning
✓	Verified: the call returned a value of the expected backend kind.
⊘	Skipped: the backend’s array-API surface does not expose the underlying op, or it rejects a keyword saiunit forwards (e.g. JAX-only precision=).
✗	Failed: the call raised an unexpected exception on this backend.
⚠	Works with a caveat (e.g. lazy result on dask, expected BackendError for materialization on dask).
🅙	JAX-only by design — gated by saiunit._jax_guard.require_jax_backend. Raises BackendError on any non-jax backend.
—	Not applicable to backend dispatch (dtype factories, dimension predicates).
?	Not tested in this report. Cupy is always ? because no CUDA backend was available when this sweep ran. The single unmapped Quantity method (tree_unflatten) is also ? because automated invocation requires a hand-crafted aux/children pair.

Sweep environment. Backends invoked: numpy, jax, torch, dask, ndonnx. Backends shown but not tested: cupy.

High-level summary

Table 1 Per-subpackage rating

Subpackage	numpy	jax	cupy	torch	dask	ndonnx
saiunit.math	Full ✓	Full ✓	?	Partial ⚠	Mostly ⚠	Partial ⚠
saiunit.linalg	Full ✓	Full ✓	?	Mostly ⚠	Partial ⚠	Limited ✗
saiunit.fft	Full ✓	Full ✓	?	Full ✓	Full ✓	Limited ✗
Quantity methods	Mostly ⚠	Mostly ⚠	?	Partial ⚠	Partial ⚠	Partial ⚠
saiunit.lax	JAX-only 🅙	Full ✓	?	JAX-only 🅙	JAX-only 🅙	JAX-only 🅙
saiunit.autograd	JAX-only 🅙	Full ✓	?	JAX-only 🅙	JAX-only 🅙	JAX-only 🅙
saiunit.sparse	JAX-only 🅙	Full ✓	?	JAX-only 🅙	JAX-only 🅙	JAX-only 🅙

Rating thresholds: Full ≥ 95 % pass and zero fail; Mostly ≥ 80 % pass; Partial ≥ 30 % pass; Limited < 30 % pass; JAX-only = gated by require_jax_backend.

Backend-specific notes

• jax — full feature set; default backend. All JAX-only subpackages (saiunit.lax, saiunit.autograd, saiunit.sparse) require this backend.

• numpy — eager CPU computation through array_api_compat.numpy. A handful of reductions (amax, amin, mean, nan* variants) fail when saiunit forwards a where=None kwarg numpy can’t interpret. These are listed with footnotes in the math tables below.

• cupy — not tested in this report (no CUDA toolkit in the sweep environment). Cells are ?. Cupy’s array-API surface tracks numpy closely, so support is expected to mirror the numpy column, but this document does not claim it.

• torch — through array_api_compat.torch. The torch array-API surface lacks several ops saiunit dispatches to (cbrt, digamma, some einops reductions, axes= for n-D FFTs) and rejects JAX-flavored kwargs (precision, symmetrize_input, tol). Affected calls are recorded as skip rather than fail.

• dask — lazy arrays. Reductions and most array ops succeed but the result remains lazy until .compute(). Per saiunit._base_quantity, the Python casts float(q) / int(q) / operator.index(q) / np.asarray(q) / hash(q) and the Quantity.tolist method raise BackendError to avoid silent materialization — these cells are ⚠ with the BackendError text in the footnote. Quantity.item on dask raises a different error (the dask Array has no .item() method) so it appears as ⊘ rather than ⚠. Methods like Quantity.float / .double are .astype in disguise and stay lazy on dask, so they pass.

• ndonnx — symbolic graph-building backend. Many array-API ops (fft.*, several linalg.*, complex / specialty math) are not implemented and surface as ⊘ skip rows. Saiunit does not encode unit information into the ONNX graph.

JAX-only subpackages

These subpackages dispatch directly to JAX primitives that have no array-API equivalent. Each entry point is wrapped with saiunit._jax_guard.require_jax_backend, which raises BackendError on any non-jax mantissa.

saiunit.lax — 101 public callable(s); all require jax.

Probe result	numpy	jax	cupy	torch	dask	ndonnx
all functions	🅙	✓	?	🅙	🅙	🅙

List of saiunit.lax functions

acos acosh approx_max_k approx_min_k asin asinh atan atan2 atanh batch_matmul bessel_i0e bessel_i1e betainc bitcast_convert_type broadcast broadcast_in_dim broadcast_shapes broadcast_to_rank broadcasted_iota cholesky clamp clz collapse complex conv conv_transpose

convert_element_type cumlogsumexp cummax cummin cumsum digamma div dot_general dynamic_index_in_dim dynamic_slice dynamic_slice_ind_dim dynamic_update_index_in_dim dynamic_update_slice dynamic_update_slice_in_dim eig eigh eq erf erf_inv erfc fft gather ge gt hessenberg

householder_product igamma igamma_grad_a igammac index_in_dim index_take integer_pow iota le lgamma logistic lt lu mul ne neg pad polygamma population_count pow qdwh qr random_gamma_grad reduce reduce_precision

rem rsqrt scatter scatter_add scatter_apply scatter_max scatter_min scatter_mul scatter_sub schur shift_left shift_right_arithmetic shift_right_logical slice slice_in_dim sort sort_key_val sub svd top_k triangular_solve tridiagonal tridiagonal_solve zeros_like_array zeta

saiunit.autograd — 7 public callable(s); all require jax.

Probe result	numpy	jax	cupy	torch	dask	ndonnx
all functions	✓	✓	?	✗	✗	✗

List of saiunit.autograd functions

grad hessian

jacfwd jacobian

jacrev value_and_grad

vector_grad

saiunit.sparse — 10 public callable(s); all require jax.

Probe result	numpy	jax	cupy	torch	dask	ndonnx
all functions	🅙	✓	?	🅙	🅙	🅙

List of saiunit.sparse functions

COO CSC CSR

SparseMatrix coo_fromdense coo_todense

csc_fromdense csc_todense

csr_fromdense csr_todense

saiunit.math

Public callables in saiunit.math that go through the multi-backend dispatcher. Grouped by source submodule for readability.

array_creation — Array creation

Table 2 saiunit.math — Array creation

Function	numpy	jax	cupy	torch	dask	ndonnx
tril_indices	✓	✓	?	✓	✓	✓
triu_indices	✓	✓	?	✓	✓	✓

keep_unit — Unit-preserving

Table 3 saiunit.math — Unit-preserving

Function	numpy	jax	cupy	torch	dask	ndonnx
astype	✓	✓	?	✓	✓	✓

change_unit — Unit-changing

saiunit.math — Unit-changing: no functions in this group.

accept_unitless — Dimensionless-only

saiunit.math — Dimensionless-only: no functions in this group.

remove_unit — Unit-removing

saiunit.math — Unit-removing: no functions in this group.

saiunit.math

Table 4 saiunit.math — saiunit.math

Function	numpy	jax	cupy	torch	dask	ndonnx
abs	✓	✓	?	✓	✓	✓
absolute	✓	✓	?	✓	✓	⊘ [1]
add	✓	✓	?	✓	✓	✓
all	✓	✓	?	✓	✓	✓
allclose	✓	✓	?	✓	✓	⊘ [2]
alltrue	✓	✓	?	✓	✓	✓
amax	✓	✓	?	✓	✓	✓
amin	✓	✓	?	✓	✓	✓
angle	✓	✓	?	✓	✓	✗ [3]
any	✓	✓	?	✓	✓	✓
append	✓	✓	?	⊘ [4]	✓	⊘ [5]
arange	✓	✓	?	✓	✓	✓
arccos	✓	✓	?	✓	✓	⊘ [6]
arccosh	✓	✓	?	✓	✓	⊘ [7]
arcsin	✓	✓	?	✓	✓	⊘ [8]
arcsinh	✓	✓	?	✓	✓	⊘ [9]
arctan	✓	✓	?	✓	✓	⊘ [10]
arctan2	✓	✓	?	✓	✓	⊘ [11]
arctanh	✓	✓	?	✓	✓	⊘ [12]
argmax	✓	✓	?	✓	✓	✓
argmin	✓	✓	?	✓	✓	✓
argsort	✓	✓	?	✓	✓	✓
argwhere	✓	✓	?	✓	✓	⊘ [13]
around	✓	✓	?	⊘ [14]	✓	⊘ [15]
array	✓	✓	?	✓	✓	✓
array_equal	✓	✓	?	⊘ [16]	⊘ [17]	⊘ [18]
array_split	✓	✓	?	✓	⊘ [19]	⊘ [20]
as_numpy	✓	✓	?	✓	✓	✓
asarray	✓	✓	?	✓	✓	✓
atleast_1d	✓	✓	?	✓	✓	⊘ [21]
atleast_2d	✓	✓	?	✓	✓	⊘ [22]
atleast_3d	✓	✓	?	✓	✓	⊘ [23]
average	✓	✓	?	⊘ [24]	✓	⊘ [25]
bincount	✓	✓	?	✓	✓	⊘ [26]
bitwise_and	✓	✓	?	✓	✓	✓
bitwise_not	✓	✓	?	✓	✓	⊘ [27]
bitwise_or	✓	✓	?	✓	✓	✓
bitwise_xor	✓	✓	?	✓	✓	✓
block	✓	✓	?	⊘ [28]	✓	⊘ [29]
broadcast_arrays	✓	✓	?	✓	✓	✓
broadcast_shapes	✓	✓	?	✓	✓	✓
broadcast_to	✓	✓	?	✓	✓	✓
cbrt	✓	✓	?	⊘ [30]	✓	⊘ [31]
ceil	✓	✓	?	✓	✓	✓
celu	✓	✓	?	🅙	🅙	🅙
choose	✓	✓	?	⊘ [32]	✓	⊘ [33]
clip	✓	✓	?	✓	✓	✓
column_stack	✓	✓	?	✓	⊘ [34]	⊘ [35]
compress	✓	✓	?	⊘ [36]	✓	⊘ [37]
concatenate	✓	✓	?	✓	✓	⊘ [38]
conj	✓	✓	?	✓	✓	✗ [3]
conjugate	✓	✓	?	⊘ [39]	⊘ [40]	✗ [3]
convolve	✓	✓	?	⊘ [41]	⊘ [42]	⊘ [43]
copysign	✓	✓	?	✓	✓	✗ [44]
corrcoef	✓	✓	?	✓	✓	⊘ [45]
correlate	✓	✓	?	⊘ [46]	⊘ [47]	⊘ [48]
cos	✓	✓	?	✓	✓	✓
cosh	✓	✓	?	✓	✓	✓
count_nonzero	✓	✓	?	✓	✓	✓
cov	✓	✓	?	✓	✓	⊘ [49]
cross	✓	✓	?	✓	⊘ [50]	⊘ [51]
cumprod	✓	✓	?	✓	✓	⊘ [52]
cumproduct	✓	✓	?	✓	✓	⊘ [52]
cumsum	✓	✓	?	✓	✓	⊘ [53]
deg2rad	✓	✓	?	✓	✓	⊘ [54]
degrees	✓	✓	?	⊘ [55]	✓	⊘ [56]
diag	✓	✓	?	✓	✓	⊘ [57]
diag_indices_from	✓	✓	?	⊘ [58]	⊘ [59]	⊘ [60]
diagflat	✓	✓	?	✓	⊘ [61]	⊘ [62]
diagonal	✓	✓	?	✓	✓	⊘ [63]
diff	✓	✓	?	✓	✓	✓
digitize	✓	✓	?	⊘ [64]	✓	⊘ [65]
divide	✓	✓	?	✓	✓	✓
divmod	✓	✓	?	⊘ [66]	✓	⊘ [67]
dot	✓	✓	?	✓	✓	⊘ [68]
dsplit	✓	✓	?	✓	⊘ [69]	⊘ [70]
dstack	✓	✓	?	✓	✓	⊘ [71]
ediff1d	✓	✓	?	⊘ [72]	✓	⊘ [73]
einsum	⊘ [74]	✓	?	⊘ [75]	⊘ [74]	⊘ [76]
elu	✓	✓	?	🅙	🅙	🅙
empty	✓	✓	?	✓	✓	✓
empty_like	✓	✓	?	✓	✓	✓
equal	✓	✓	?	✓	✓	✓
exp	✓	✓	?	✓	✓	✓
exp2	✓	✓	?	✓	✓	⊘ [77]
expand_dims	✓	✓	?	✓	✓	✓
expm1	✓	✓	?	✓	✓	✗ [44]
exprel	✓	✓	?	🅙	🅙	🅙
extract	✓	✓	?	⊘ [78]	✓	⊘ [79]
eye	✓	✓	?	✓	✓	✓
fabs	✓	✓	?	⊘ [80]	✓	⊘ [81]
fill_diagonal	✓	✓	?	⊘ [82]	⊘ [83]	⊘ [84]
finfo	✓	✓	?	✓	✓	✓
fix	✓	✓	?	✓	✓	✓
flatnonzero	✓	✓	?	⊘ [85]	✓	⊘ [86]
flatten	✓	✓	?	✓	✓	✓
flip	✓	✓	?	✓	✓	✓
fliplr	✓	✓	?	✓	✓	⊘ [87]
flipud	✓	✓	?	✓	✓	⊘ [88]
float_power	✓	✓	?	✓	✓	⊘ [89]
floor	✓	✓	?	✓	✓	✓
floor_divide	✓	✓	?	✓	✓	✓
fmax	✓	✓	?	✓	✓	⊘ [90]
fmin	✓	✓	?	✓	✓	⊘ [91]
fmod	✓	✓	?	✓	✓	⊘ [92]
frexp	✓	✓	?	✓	✓	⊘ [93]
from_numpy	✓	✓	?	✓	✓	✓
full	✓	✓	?	✓	✓	✓
full_like	✓	✓	?	✓	✓	✓
gather	✓	✓	?	✓	✗ [94]	⊘ [95]
gcd	✓	✓	?	✓	⊘ [96]	⊘ [97]
gelu	✓	✓	?	🅙	🅙	🅙
get_promote_dtypes	✓	✓	?	✓	✓	✓
glu	✓	✓	?	🅙	🅙	🅙
gradient	✓	✓	?	✓	✓	⊘ [98]
greater	✓	✓	?	✓	✓	✓
greater_equal	✓	✓	?	✓	✓	✓
hard_sigmoid	✓	✓	?	🅙	🅙	🅙
hard_silu	✓	✓	?	🅙	🅙	🅙
hard_swish	✓	✓	?	🅙	🅙	🅙
hard_tanh	✓	✓	?	🅙	🅙	🅙
heaviside	✓	✓	?	✓	⊘ [99]	⊘ [100]
histogram	✓	✓	?	✓	✓	⊘ [101]
hsplit	✓	✓	?	✓	⊘ [102]	⊘ [103]
hstack	✓	✓	?	✓	✓	⊘ [104]
hypot	✓	✓	?	✓	✓	✗ [44]
identity	✓	✓	?	⊘ [105]	⊘ [106]	⊘ [107]
iinfo	✓	✓	?	✓	✓	✓
imag	✓	✓	?	✓	✓	✗ [3]
inner	✓	✓	?	✓	⊘ [108]	⊘ [109]
interp	✓	✓	?	⊘ [110]	⊘ [111]	⊘ [112]
intersect1d	✓	✓	?	⊘ [113]	⊘ [114]	⊘ [115]
invert	✓	✓	?	⊘ [116]	✓	⊘ [117]
isclose	✓	✓	?	✓	✓	⊘ [118]
iscomplexobj	✓	✓	?	⊘ [119]	⊘ [120]	⊘ [121]
isfinite	✓	✓	?	✓	✓	✓
isinf	✓	✓	?	✓	✓	✓
isnan	✓	✓	?	✓	✓	✓
isreal	✓	✓	?	✓	✓	⊘ [122]
isscalar	✓	✓	?	✓	✓	✓
issubdtype	✓	✓	?	✓	✓	✓
kron	✓	✓	?	✓	⊘ [123]	⊘ [124]
lcm	✓	✓	?	✓	⊘ [125]	⊘ [126]
ldexp	✓	✓	?	✓	✓	⊘ [127]
leaky_relu	✓	✓	?	✓	✓	✓
left_shift	✓	✓	?	⊘ [128]	✓	⊘ [129]
less	✓	✓	?	✓	✓	✓
less_equal	✓	✓	?	✓	✓	✓
linspace	✓	✓	?	✓	✓	✓
log	✓	✓	?	✓	✓	✓
log10	✓	✓	?	✓	✓	✓
log1p	✓	✓	?	✓	✓	✗ [44]
log2	✓	✓	?	✓	✓	✓
log_sigmoid	✓	✓	?	🅙	🅙	🅙
logaddexp	✓	✓	?	✓	✓	✓
logaddexp2	✓	✓	?	✓	✓	⊘ [130]
logical_and	✓	✓	?	✓	✓	✓
logical_not	✓	✓	?	✓	✓	✓
logical_or	✓	✓	?	✓	✓	✓
logical_xor	✓	✓	?	✓	✓	✓
logspace	✓	✓	?	⊘ [131]	⊘ [132]	⊘ [133]
matmul	✓	✓	?	✓	✓	✓
matrix_power	✓	✓	?	✓	⊘ [134]	⊘ [135]
max	✓	✓	?	✓	✓	✓
maximum	✓	✓	?	✓	✓	✓
mean	✓	✓	?	✓	✓	✓
median	✓	✓	?	✓	✗ [136]	⊘ [137]
meshgrid	✓	✓	?	✓	✓	⊘ [138]
min	✓	✓	?	✓	✓	✓
minimum	✓	✓	?	✓	✓	✓
mish	✓	✓	?	🅙	🅙	🅙
mod	✓	✓	?	⊘ [139]	✓	⊘ [140]
modf	✓	✓	?	⊘ [141]	✓	⊘ [142]
moveaxis	✓	✓	?	✓	✓	✓
multi_dot	✓	✓	?	✓	⊘ [143]	⊘ [144]
multiply	✓	✓	?	✓	✓	✓
nan_to_num	✓	✓	?	✓	✓	⊘ [145]
nanargmax	✓	✓	?	⊘ [146]	✓	⊘ [147]
nanargmin	✓	✓	?	⊘ [148]	✓	⊘ [149]
nancumprod	✓	✓	?	⊘ [150]	✓	⊘ [151]
nancumsum	✓	✓	?	⊘ [152]	✓	⊘ [153]
nanmax	✓	✓	?	⊘ [154]	✓	⊘ [155]
nanmean	✓	✓	?	✓	✓	⊘ [156]
nanmedian	✓	✓	?	✓	✗ [157]	⊘ [158]
nanmin	✓	✓	?	⊘ [159]	✓	⊘ [160]
nanpercentile	✓	✓	?	⊘ [161]	✓	⊘ [162]
nanprod	✓	✓	?	⊘ [163]	✓	⊘ [164]
nanquantile	✓	✓	?	⊘ [165]	✓	⊘ [166]
nanstd	✓	✓	?	⊘ [167]	✓	⊘ [168]
nansum	✓	✓	?	✓	✓	⊘ [169]
nanvar	✓	✓	?	⊘ [170]	✓	⊘ [171]
ndim	✓	✓	?	✓	✓	✓
negative	✓	✓	?	✓	✓	✓
nextafter	✓	✓	?	✓	✓	✗ [44]
nonzero	✓	✓	?	✓	✓	✓
not_equal	✓	✓	?	✓	✓	✓
ones	✓	✓	?	✓	✓	✓
ones_like	✓	✓	?	✓	✓	✓
outer	✓	✓	?	✓	✓	⊘ [172]
percentile	✓	✓	?	⊘ [173]	✓	⊘ [174]
positive	✓	✓	?	✓	✓	✓
power	✓	✓	?	⊘ [175]	✓	⊘ [176]
prod	✓	✓	?	✓	✓	✓
product	✓	✓	?	✓	✓	✓
promote_dtypes	✓	✓	?	✓	✓	✓
ptp	✓	✓	?	⊘ [177]	✓	⊘ [178]
quantile	✓	✓	?	⊘ [179]	✓	⊘ [180]
rad2deg	✓	✓	?	✓	✓	⊘ [181]
radians	✓	✓	?	⊘ [182]	✓	⊘ [183]
ravel	✓	✓	?	✓	✓	⊘ [184]
real	✓	✓	?	✓	✓	✗ [3]
reciprocal	✓	✓	?	✓	✓	✓
relu	✓	✓	?	🅙	🅙	🅙
relu6	✓	✓	?	🅙	🅙	🅙
remainder	✓	✓	?	✓	✓	✓
remove_diag	✓	✓	?	✓	✓	⊘ [185]
repeat	✓	✓	?	✓	✓	✓
reshape	✓	✓	?	✓	✓	✓
result_type	✓	✓	?	✓	✓	✓
right_shift	✓	✓	?	⊘ [186]	✓	⊘ [187]
rint	✓	✓	?	⊘ [188]	✓	⊘ [189]
roll	✓	✓	?	✓	✓	✓
rot90	✓	✓	?	✓	✓	⊘ [190]
round	✓	✓	?	✓	✓	✓
row_stack	✓	✓	?	✓	✓	⊘ [191]
searchsorted	✓	✓	?	✓	✓	✓
select	✓	✓	?	⊘ [192]	✓	⊘ [193]
selu	✓	✓	?	🅙	🅙	🅙
shape	✓	✓	?	✓	✓	✓
sigmoid	✓	✓	?	🅙	🅙	🅙
sign	✓	✓	?	✓	✓	✓
signbit	✓	✓	?	✓	✓	✗ [44]
silu	✓	✓	?	🅙	🅙	🅙
sin	✓	✓	?	✓	✓	✓
sinc	✓	✓	?	✓	✓	⊘ [194]
sinh	✓	✓	?	✓	✓	✓
size	✓	✓	?	✓	✓	✓
soft_sign	✓	✓	?	🅙	🅙	🅙
softplus	✓	✓	?	🅙	🅙	🅙
sometrue	✓	✓	?	✓	✓	✓
sort	✓	✓	?	✓	✓	✓
sparse_plus	✓	✓	?	🅙	🅙	🅙
sparse_sigmoid	✓	✓	?	🅙	🅙	🅙
split	✓	✓	?	✓	⊘ [19]	⊘ [20]
sqrt	✓	✓	?	✓	✓	✓
square	✓	✓	?	✓	✓	✓
squareplus	✓	✓	?	🅙	🅙	🅙
squeeze	✓	✓	?	⊘ [195]	✓	⊘ [195]
stack	✓	✓	?	✓	✓	✓
std	✓	✓	?	✓	✓	✓
subtract	✓	✓	?	✓	✓	✓
sum	✓	✓	?	✓	✓	✓
swapaxes	✓	✓	?	✓	✓	⊘ [196]
swish	✓	✓	?	🅙	🅙	🅙
take	✓	✓	?	✓	✓	✓
tan	✓	✓	?	✓	✓	✓
tanh	✓	✓	?	✓	✓	✓
tensordot	✓	✓	?	✓	✓	✓
tile	✓	✓	?	✓	✓	✓
trace	✓	✓	?	✓	✓	⊘ [197]
transpose	✓	✓	?	⊘ [198]	✓	⊘ [199]
tree_ones_like	✓	✓	?	✓	✓	✓
tree_zeros_like	✓	✓	?	✓	✓	✓
tri	✓	✓	?	⊘ [200]	✓	⊘ [201]
tril	✓	✓	?	✓	✓	✓
tril_indices_from	✓	✓	?	⊘ [202]	✓	⊘ [203]
triu	✓	✓	?	✓	✓	✓
triu_indices_from	✓	✓	?	⊘ [204]	✓	⊘ [205]
true_divide	✓	✓	?	✓	✓	⊘ [206]
trunc	✓	✓	?	✓	✓	✓
unflatten	✓	✓	?	✓	✓	✓
unique	✓	✓	?	✓	✓	⊘ [207]
vander	✓	✓	?	✓	⊘ [208]	⊘ [209]
var	✓	✓	?	✓	✓	✓
vdot	✓	✓	?	✓	✓	⊘ [210]
vecdot	✓	✓	?	✓	✓	✓
vsplit	✓	✓	?	✓	⊘ [211]	⊘ [212]
vstack	✓	✓	?	✓	✓	⊘ [191]
where	✓	✓	?	✓	✓	✓
zeros	✓	✓	?	✓	✓	✓
zeros_like	✓	✓	?	✓	✓	✓

jax.numpy

Table 5 saiunit.math — jax.numpy

Function	numpy	jax	cupy	torch	dask	ndonnx
bartlett	✓	✓	?	✓	✓	✓
blackman	✓	✓	?	✓	✓	✓
hamming	✓	✓	?	✓	✓	✓
hanning	✓	✓	?	✓	✓	✓
kaiser	✓	✓	?	✓	✓	✓

numpy

Table 6 saiunit.math — numpy

Function	numpy	jax	cupy	torch	dask	ndonnx
dtype	✓	✓	?	✓	✓	✓

brainstate.math

Table 7 saiunit.math — brainstate.math

Function	numpy	jax	cupy	torch	dask	ndonnx
einrearrange	✓	✓	?	⊘ [213]	✓	⊘ [199]
einreduce	✓	✓	?	✓	✓	✓
einrepeat	✓	✓	?	✓	✓	✓
einshape	✓	✓	?	✓	✓	✓

Non-dispatched helpers

These names live under saiunit.math for convenience but do not dispatch on backend — they are dtype factories (re-exported from jax.numpy) or pure-Python predicates / introspection helpers over Quantity / Unit objects. Behavior is identical on every backend.

assert_quantity bfloat16 bool_ cdouble check_dims check_units complex128 complex64 complex_ csingle display_in_unit double

fail_for_dimension_mismatch fail_for_unit_mismatch float16 float32 float64 float_ get_dim get_dtype get_magnitude get_mantissa get_or_create_dimension get_unit

inexact int16 int2 int32 int4 int64 int8 int_ is_dimensionless is_float is_int is_quantity

is_unitless maybe_decimal set_exprel_order single uint uint16 uint2 uint32 uint4 uint64 uint8

saiunit.linalg

Table 8 saiunit.linalg

Function	numpy	jax	cupy	torch	dask	ndonnx
cholesky	✓	✓	?	✓	✓	⊘ [214]
cond	✓	✓	?	✓	⊘ [215]	⊘ [216]
cross	✓	✓	?	✓	⊘ [50]	⊘ [51]
det	✓	✓	?	✓	⊘ [217]	⊘ [218]
diagonal	✓	✓	?	✓	✓	⊘ [63]
dot	✓	✓	?	✓	✓	⊘ [68]
eig	✓	✓	?	⊘ [219]	⊘ [220]	⊘ [221]
eigh	✓	✓	?	⊘ [222]	⊘ [223]	⊘ [224]
eigvals	✓	✓	?	⊘ [219]	⊘ [220]	⊘ [221]
eigvalsh	✓	✓	?	⊘ [222]	⊘ [223]	⊘ [224]
inner	✓	✓	?	✓	⊘ [108]	⊘ [109]
inv	✓	✓	?	✓	✓	⊘ [225]
kron	✓	✓	?	✓	⊘ [123]	⊘ [124]
lstsq	✓	✓	?	✓	⊘ [226]	⊘ [227]
matmul	✓	✓	?	✓	✓	✓
matrix_norm	✓	✓	?	✓	✓	⊘ [228]
matrix_power	✓	✓	?	✓	⊘ [134]	⊘ [135]
matrix_rank	✓	✓	?	✓	⊘ [229]	⊘ [230]
matrix_transpose	✓	✓	?	✓	✓	⊘ [231]
multi_dot	✓	✓	?	✓	⊘ [143]	⊘ [144]
norm	✓	✓	?	✓	✓	⊘ [232]
outer	✓	✓	?	✓	✓	⊘ [172]
pinv	✓	✓	?	✓	⊘ [233]	⊘ [234]
qr	✓	✓	?	✓	✓	⊘ [227]
slogdet	✓	✓	?	✓	⊘ [235]	⊘ [227]
solve	✓	✓	?	✓	✓	⊘ [236]
svd	✓	✓	?	⊘ [237]	⊘ [238]	⊘ [239]
svdvals	✓	✓	?	⊘ [237]	⊘ [238]	⊘ [239]
tensordot	✓	✓	?	✓	✓	✓
tensorinv	✓	✓	?	✓	⊘ [240]	⊘ [241]
tensorsolve	✓	✓	?	✓	⊘ [242]	⊘ [243]
trace	✓	✓	?	✓	✓	⊘ [197]
vdot	✓	✓	?	✓	✓	⊘ [210]
vecdot	✓	✓	?	✓	✓	✓
vector_norm	✓	✓	?	✓	✓	⊘ [244]

saiunit.fft

Routing varies inside saiunit.fft: _fft_change_unit.py calls saiunit._backend.get_backend() directly (e.g. for fftfreq / rfftfreq), while _fft_keep_unit.py delegates to the math package’s _fun_keep_unit_unary helper and inherits its dispatch.

Table 9 saiunit.fft

Function	numpy	jax	cupy	torch	dask	ndonnx
fft	✓	✓	?	✓	✓	⊘ [245]
fft2	✓	✓	?	✓	✓	⊘ [246]
fftfreq	✓	✓	?	✓	✓	⊘ [247]
fftn	✓	✓	?	✓	✓	⊘ [248]
fftshift	✓	✓	?	✓	✓	⊘ [249]
ifft	✓	✓	?	✓	✓	⊘ [250]
ifft2	✓	✓	?	✓	✓	⊘ [251]
ifftn	✓	✓	?	✓	✓	⊘ [252]
ifftshift	✓	✓	?	✓	✓	⊘ [253]
irfft	✓	✓	?	✓	✓	⊘ [254]
irfft2	✓	✓	?	✓	✓	⊘ [255]
irfftn	✓	✓	?	✓	✓	⊘ [256]
rfft	✓	✓	?	✓	✓	⊘ [257]
rfft2	✓	✓	?	✓	✓	⊘ [258]
rfftfreq	✓	✓	?	✓	✓	⊘ [247]
rfftn	✓	✓	?	✓	✓	⊘ [259]

Quantity methods

Methods on saiunit.Quantity itself. .to_<backend>() methods ignore the current backend and convert to the named one — cells show ⊘ when the target backend isn’t installed in the sweep environment.

Materialization is documented above (see Backend-specific notes). Quantity.tolist on dask is the one method that raises BackendError from saiunit’s own guard (⚠). .item reports ⊘ on dask / ndonnx because the underlying array object does not expose .item().

Table 10 Quantity public methods

Function	numpy	jax	cupy	torch	dask	ndonnx
all	✓	✓	?	✓	✓	✓
any	✓	✓	?	✓	✓	✓
argmax	✓	✓	?	✓	✓	✓
argmin	✓	✓	?	✓	✓	✓
argsort	✓	✓	?	✓	✓	✓
astype	✓	✓	?	✓	✓	✓
clip	✓	✓	?	✓	✓	✓
clone	✓	✓	?	⊘ [260]	⊘ [261]	⊘ [262]
conj	✓	✓	?	✓	✓	✗ [44]
conjugate	✓	✓	?	✓	✓	✗ [44]
copy	✓	✓	?	⊘ [260]	⊘ [261]	⊘ [262]
cpu	✓	✓	?	✗ [263]	✗ [264]	✗ [265]
cross	✓	✓	?	⊘ [266]	⊘ [267]	⊘ [268]
cuda	⊘ [269]	⊘ [269]	?	⊘ [269]	⊘ [269]	⊘ [269]
cumprod	⊘ [270]	⊘ [270]	?	⊘ [270]	⊘ [270]	⊘ [270]
cumsum	✓	✓	?	⊘ [271]	✓	⊘ [53]
diagonal	✓	✓	?	⊘ [272]	✓	⊘ [273]
dot	✓	✓	?	✓	✓	⊘ [274]
double	✓	✓	?	✓	✓	✓
expand_as	✓	✓	?	✓	✓	✓
expand_dims	✓	✓	?	✓	✓	✓
factorless	✓	✓	?	✓	✓	✓
fill	✓	✓	?	✗ [263]	✗ [275]	✗ [276]
flatten	✓	✓	?	✓	✓	✓
float	✓	✓	?	✓	✓	✓
half	✓	✓	?	✓	⊘ [277]	✓
has_same_unit	✓	✓	?	✓	✓	✓
in_unit	✓	✓	?	✓	✓	✓
item	✓	✓	?	✓	⊘ [278]	⊘ [278]
max	✓	✓	?	✓	✓	✓
mean	✓	✓	?	✓	✓	✓
min	✓	✓	?	✓	✓	✓
nancumprod	⊘ [279]	⊘ [279]	?	⊘ [279]	⊘ [279]	⊘ [279]
nanprod	✓	✓	?	⊘ [163]	✓	⊘ [164]
nonzero	✓	✓	?	✓	✓	✓
outer	✓	✓	?	✓	✓	⊘ [280]
pow	✓	✓	?	✓	✓	✓
prod	✓	✓	?	✓	✓	✓
ptp	✓	✓	?	⊘ [177]	✓	⊘ [178]
put	✓	✓	?	✗ [263]	⊘ [281]	✗ [276]
ravel	✓	✓	?	✓	✓	⊘ [184]
repeat	✓	✓	?	✓	✓	✓
repr_in_unit	✓	✓	?	✓	✓	✓
reshape	✓	✓	?	✓	✓	✓
resize	✓	✓	?	✗ [263]	✗ [275]	✗ [282]
round	✓	✓	?	✓	✓	✓
scatter_add	✓	✓	?	✓	⊘ [283]	✗ [276]
scatter_div	✓	✓	?	✓	⊘ [284]	✗ [276]
scatter_max	✓	✓	?	✓	⊘ [285]	✗ [276]
scatter_min	✓	✓	?	✓	⊘ [286]	✗ [276]
scatter_mul	✓	✓	?	✓	⊘ [287]	✗ [276]
scatter_sub	✓	✓	?	✓	⊘ [283]	✗ [276]
searchsorted	✓	✓	?	✓	✓	✓
sort	✓	✓	?	✗ [263]	✗ [275]	✗ [265]
split	✓	✓	?	⊘ [288]	✓	✗ [289]
squeeze	✓	✓	?	✗ [290]	✓	✗ [291]
std	✓	✓	?	✓	✓	✓
sum	✓	✓	?	✓	✓	✓
swapaxes	✓	✓	?	✓	✓	⊘ [292]
take	✓	✓	?	✓	✗ [293]	✓
tile	✓	✓	?	✗ [294]	✓	✗ [295]
to	✓	✓	?	✓	✓	✓
to_cupy	⊘ [296]	⊘ [296]	?	⊘ [296]	⊘ [296]	⊘ [296]
to_dask	✓	✓	?	✗ [263]	✓	✗ [265]
to_decimal	✓	✓	?	✓	✓	✓
to_jax	✓	✓	?	✓	✓	✗ [297]
to_ndonnx	✓	⊘ [298]	?	⊘ [299]	⊘ [300]	✓
to_numpy	✓	✓	?	✓	✓	✓
to_torch	✓	✓	?	✓	✗ [301]	✗ [302]
tolist	✓	✓	?	✓	⚠ [303]	⊘ [304]
trace	✓	✓	?	⊘ [305]	✓	⊘ [306]
transpose	✓	✓	?	⊘ [198]	✓	⊘ [307]
tree_flatten	✓	✓	?	✓	✓	✓
tree_unflatten	?	?	?	?	?	?
unsqueeze	✓	✓	?	✓	✓	✓
update_mantissa	✓	✓	?	✗ [263]	✗ [275]	✗ [265]
var	✓	✓	?	✓	✓	✓
view	✓	✓	?	⊘ [308]	✓	⊘ [309]
with_unit	✓	✓	?	✓	✓	✓

Coverage statistic

Subpackage	Mapped	Non-dispatched	Unmapped	Total
saiunit.math	294	47	0	341
saiunit.linalg	35	0	0	35
saiunit.fft	16	0	0	16
Quantity	78	0	1	79

Mapped = functions the sweep actually invoked. Non-dispatched = type factories / predicates that don’t go through backend dispatch. Unmapped = no call pattern registered (will appear as ? in tables).

How this was produced

dev/backend_support_sweep.py walks every public callable in the subpackages above, picks a calling pattern from an in-script registry, and invokes the function under with saiunit.using_backend(b) for each backend b in the local environment. Outcomes are classified as pass / skip / fail / warn / unmapped / na and written to dev/backend_support_data.json.

dev/backend_support_render.py (this script’s source) reads that JSON and emits the rst file you are currently reading. To refresh:

PYTHONPATH=. python dev/backend_support_sweep.py
PYTHONPATH=. python dev/backend_support_render.py

JAX-only subpackages are probed with one representative function per subpackage rather than enumerated — every entry point in saiunit.lax / .autograd / .sparse is gated identically.

Footnotes

[1]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘absolute’

[2]

AttributeError: module ‘ndonnx’ has no attribute allclose

[3] (1,2,3,4,5)

ValueError: ‘complex128’ does not have a corresponding ndonnx data type

[4]

AttributeError: saiunit: backend ‘array_api_compat.torch’ has no operation ‘append’

[5]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘append’

[6]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘arccos’

[7]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘arccosh’

[8]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘arcsin’

[9]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘arcsinh’

[10]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘arctan’

[11]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘arctan2’

[12]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘arctanh’

[13]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘argwhere’

[14]

AttributeError: saiunit: backend ‘array_api_compat.torch’ has no operation ‘around’

[15]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘around’

[16]

AttributeError: saiunit: backend ‘array_api_compat.torch’ has no operation ‘array_equal’

[17]

AttributeError: saiunit: backend ‘array_api_compat.dask.array’ has no operation ‘array_equal’

[18]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘array_equal’

[19] (1,2)

AttributeError: saiunit: backend ‘array_api_compat.dask.array’ has no operation ‘split’

[20] (1,2)

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘split’

[21]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘atleast_1d’

[22]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘atleast_2d’

[23]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘atleast_3d’

[24]

AttributeError: saiunit: backend ‘array_api_compat.torch’ has no operation ‘average’

[25]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘average’

[26]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘bincount’

[27]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘bitwise_not’

[28]

AttributeError: saiunit: backend ‘array_api_compat.torch’ has no operation ‘block’

[29]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘block’

[30]

AttributeError: saiunit: backend ‘array_api_compat.torch’ has no operation ‘cbrt’

[31]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘cbrt’

[32]

AttributeError: saiunit: backend ‘array_api_compat.torch’ has no operation ‘choose’

[33]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘choose’

[34]

AttributeError: saiunit: backend ‘array_api_compat.dask.array’ has no operation ‘column_stack’

[35]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘column_stack’

[36]

AttributeError: saiunit: backend ‘array_api_compat.torch’ has no operation ‘compress’

[37]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘compress’

[38]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘concatenate’

[39]

AttributeError: saiunit: backend ‘array_api_compat.torch’ has no operation ‘conjugate’

[40]

AttributeError: saiunit: backend ‘array_api_compat.dask.array’ has no operation ‘conjugate’

[41]

AttributeError: saiunit: backend ‘array_api_compat.torch’ has no operation ‘convolve’

[42]

AttributeError: saiunit: backend ‘array_api_compat.dask.array’ has no operation ‘convolve’

[43]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘convolve’

[44] (1,2,3,4,5,6,7,8)

NotImplementedError:

[45]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘corrcoef’

[46]

AttributeError: saiunit: backend ‘array_api_compat.torch’ has no operation ‘correlate’

[47]

AttributeError: saiunit: backend ‘array_api_compat.dask.array’ has no operation ‘correlate’

[48]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘correlate’

[49]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘cov’

[50] (1,2)

AttributeError: saiunit: backend ‘array_api_compat.dask.array’ has no operation ‘cross’

[51] (1,2)

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘cross’

[52] (1,2)

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘cumprod’

[53] (1,2)

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘cumsum’

[54]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘deg2rad’

[55]

AttributeError: saiunit: backend ‘array_api_compat.torch’ has no operation ‘degrees’

[56]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘degrees’

[57]

AttributeError: module ‘ndonnx’ has no attribute diag

[58]

AttributeError: saiunit: backend ‘array_api_compat.torch’ has no operation ‘diag_indices_from’

[59]

AttributeError: saiunit: backend ‘array_api_compat.dask.array’ has no operation ‘diag_indices_from’

[60]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘diag_indices_from’

[61]

AttributeError: saiunit: backend ‘array_api_compat.dask.array’ has no operation ‘diagflat’

[62]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘diagflat’

[63] (1,2)

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘diagonal’

[64]

AttributeError: saiunit: backend ‘array_api_compat.torch’ has no operation ‘digitize’

[65]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘digitize’

[66]

AttributeError: saiunit: backend ‘array_api_compat.torch’ has no operation ‘divmod’

[67]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘divmod’

[68] (1,2)

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘dot’

[69]

AttributeError: saiunit: backend ‘array_api_compat.dask.array’ has no operation ‘dsplit’

[70]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘dsplit’

[71]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘dstack’

[72]

AttributeError: saiunit: backend ‘array_api_compat.torch’ has no operation ‘ediff1d’

[73]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘ediff1d’

[74] (1,2)

TracerArrayConversionError: The numpy.ndarray conversion method __array__() was called on traced array with shape float32[2,2] The error occurred while tracing the function _einsum at /mnt/d/codes/…

[75]

TypeError: Error interpreting argument to <function _einsum at 0x…> as an abstract array. The problematic value is of type <class ‘torch.Tensor’> and was passed to the function at path operands[0…

[76]

TypeError: Error interpreting argument to <function _einsum at 0x…> as an abstract array. The problematic value is of type <class ‘ndonnx._array.Array’> and was passed to the function at path ope…

[77]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘exp2’

[78]

AttributeError: saiunit: backend ‘array_api_compat.torch’ has no operation ‘extract’

[79]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘extract’

[80]

AttributeError: saiunit: backend ‘array_api_compat.torch’ has no operation ‘fabs’

[81]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘fabs’

[82]

AttributeError: module ‘array_api_compat.torch’ has no attribute ‘fill_diagonal’

[83]

AttributeError: module ‘array_api_compat.dask.array’ has no attribute ‘fill_diagonal’

[84]

AttributeError: module ‘ndonnx’ has no attribute fill_diagonal

[85]

AttributeError: saiunit: backend ‘array_api_compat.torch’ has no operation ‘flatnonzero’

[86]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘flatnonzero’

[87]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘fliplr’

[88]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘flipud’

[89]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘float_power’

[90]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘fmax’

[91]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘fmin’

[92]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘fmod’

[93]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘frexp’

[94]

NotImplementedError: Don’t yet support nd fancy indexing

[95]

AttributeError: ‘Array’ object has no attribute ‘reshape’

[96]

AttributeError: saiunit: backend ‘array_api_compat.dask.array’ has no operation ‘gcd’

[97]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘gcd’

[98]

AttributeError: module ‘ndonnx’ has no attribute gradient

[99]

AttributeError: saiunit: backend ‘array_api_compat.dask.array’ has no operation ‘heaviside’

[100]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘heaviside’

[101]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘histogram’

[102]

AttributeError: saiunit: backend ‘array_api_compat.dask.array’ has no operation ‘hsplit’

[103]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘hsplit’

[104]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘hstack’

[105]

AttributeError: module ‘array_api_compat.torch’ has no attribute ‘identity’

[106]

AttributeError: module ‘array_api_compat.dask.array’ has no attribute ‘identity’

[107]

AttributeError: module ‘ndonnx’ has no attribute identity

[108] (1,2)

AttributeError: saiunit: backend ‘array_api_compat.dask.array’ has no operation ‘inner’

[109] (1,2)

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘inner’

[110]

AttributeError: saiunit: backend ‘array_api_compat.torch’ has no operation ‘interp’

[111]

AttributeError: saiunit: backend ‘array_api_compat.dask.array’ has no operation ‘interp’

[112]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘interp’

[113]

AttributeError: saiunit: backend ‘array_api_compat.torch’ has no operation ‘intersect1d’

[114]

AttributeError: saiunit: backend ‘array_api_compat.dask.array’ has no operation ‘intersect1d’

[115]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘intersect1d’

[116]

AttributeError: saiunit: backend ‘array_api_compat.torch’ has no operation ‘invert’

[117]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘invert’

[118]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘isclose’

[119]

AttributeError: saiunit: backend ‘array_api_compat.torch’ has no operation ‘iscomplexobj’

[120]

AttributeError: saiunit: backend ‘array_api_compat.dask.array’ has no operation ‘iscomplexobj’

[121]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘iscomplexobj’

[122]

AttributeError: module ‘ndonnx’ has no attribute isreal

[123] (1,2)

AttributeError: saiunit: backend ‘array_api_compat.dask.array’ has no operation ‘kron’

[124] (1,2)

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘kron’

[125]

AttributeError: saiunit: backend ‘array_api_compat.dask.array’ has no operation ‘lcm’

[126]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘lcm’

[127]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘ldexp’

[128]

AttributeError: saiunit: backend ‘array_api_compat.torch’ has no operation ‘left_shift’

[129]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘left_shift’

[130]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘logaddexp2’

[131]

TypeError: logspace() received an invalid combination of arguments - got (float, float), but expected one of: * (Tensor start, Tensor end, int steps, float base = 10.0, *, Tensor out = None, torch….

[132]

AttributeError: module ‘array_api_compat.dask.array’ has no attribute ‘logspace’

[133]

AttributeError: module ‘ndonnx’ has no attribute logspace

[134] (1,2)

AttributeError: saiunit: backend ‘array_api_compat.dask.array’ has no operation ‘linalg.matrix_power’

[135] (1,2)

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘linalg.matrix_power’

[136]

NotImplementedError: The da.median function only works along an axis. The full algorithm is difficult to do in parallel

[137]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘median’

[138]

AttributeError: ‘list’ object has no attribute ‘ndim’

[139]

AttributeError: saiunit: backend ‘array_api_compat.torch’ has no operation ‘mod’

[140]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘mod’

[141]

AttributeError: saiunit: backend ‘array_api_compat.torch’ has no operation ‘modf’

[142]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘modf’

[143] (1,2)

AttributeError: saiunit: backend ‘array_api_compat.dask.array’ has no operation ‘linalg.multi_dot’

[144] (1,2)

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘linalg.multi_dot’

[145]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘nan_to_num’

[146]

AttributeError: saiunit: backend ‘array_api_compat.torch’ has no operation ‘nanargmax’

[147]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘nanargmax’

[148]

AttributeError: saiunit: backend ‘array_api_compat.torch’ has no operation ‘nanargmin’

[149]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘nanargmin’

[150]

AttributeError: saiunit: backend ‘array_api_compat.torch’ has no operation ‘nancumprod’

[151]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘nancumprod’

[152]

AttributeError: saiunit: backend ‘array_api_compat.torch’ has no operation ‘nancumsum’

[153]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘nancumsum’

[154]

AttributeError: saiunit: backend ‘array_api_compat.torch’ has no operation ‘nanmax’

[155]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘nanmax’

[156]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘nanmean’

[157]

NotImplementedError: The da.nanmedian function only works along an axis or a subset of axes. The full algorithm is difficult to do in parallel

[158]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘nanmedian’

[159]

AttributeError: saiunit: backend ‘array_api_compat.torch’ has no operation ‘nanmin’

[160]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘nanmin’

[161]

AttributeError: saiunit: backend ‘array_api_compat.torch’ has no operation ‘nanpercentile’

[162]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘nanpercentile’

[163] (1,2)

AttributeError: saiunit: backend ‘array_api_compat.torch’ has no operation ‘nanprod’

[164] (1,2)

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘nanprod’

[165]

TypeError: nanquantile() received an invalid combination of arguments - got (Tensor), but expected one of: * (Tensor input, Tensor q, int dim = None, bool keepdim = False, *, str interpolation = “l…

[166]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘nanquantile’

[167]

AttributeError: saiunit: backend ‘array_api_compat.torch’ has no operation ‘nanstd’

[168]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘nanstd’

[169]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘nansum’

[170]

AttributeError: saiunit: backend ‘array_api_compat.torch’ has no operation ‘nanvar’

[171]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘nanvar’

[172] (1,2)

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘outer’

[173]

AttributeError: saiunit: backend ‘array_api_compat.torch’ has no operation ‘percentile’

[174]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘percentile’

[175]

AttributeError: saiunit: backend ‘array_api_compat.torch’ has no operation ‘power’

[176]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘power’

[177] (1,2)

AttributeError: saiunit: backend ‘array_api_compat.torch’ has no operation ‘ptp’

[178] (1,2)

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘ptp’

[179]

TypeError: quantile() received an invalid combination of arguments - got (Tensor), but expected one of: * (Tensor input, Tensor q, int dim = None, bool keepdim = False, *, str interpolation = “line…

[180]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘quantile’

[181]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘rad2deg’

[182]

AttributeError: saiunit: backend ‘array_api_compat.torch’ has no operation ‘radians’

[183]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘radians’

[184] (1,2)

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘ravel’

[185]

AttributeError: type object ‘bool’ has no attribute ‘__ndx_create__’

[186]

AttributeError: saiunit: backend ‘array_api_compat.torch’ has no operation ‘right_shift’

[187]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘right_shift’

[188]

AttributeError: saiunit: backend ‘array_api_compat.torch’ has no operation ‘rint’

[189]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘rint’

[190]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘rot90’

[191] (1,2)

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘vstack’

[192]

TypeError: select() received an invalid combination of arguments - got (list, list), but expected one of: * (Tensor input, name dim, int index) * (Tensor input, int dim, int index)

[193]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘select’

[194]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘sinc’

[195] (1,2)

TypeError: squeeze() missing 1 required positional argument: ‘axis’

[196]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘swapaxes’

[197] (1,2)

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘trace’

[198] (1,2)

TypeError: transpose() received an invalid combination of arguments - got (Tensor), but expected one of: * (Tensor input, int dim0, int dim1) * (Tensor input, name dim0, name dim1)

[199] (1,2)

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘transpose’

[200]

AttributeError: module ‘array_api_compat.torch’ has no attribute ‘tri’

[201]

AttributeError: module ‘ndonnx’ has no attribute tri

[202]

AttributeError: module ‘array_api_compat.torch’ has no attribute ‘tril_indices_from’

[203]

AttributeError: module ‘ndonnx’ has no attribute tril_indices_from

[204]

AttributeError: module ‘array_api_compat.torch’ has no attribute ‘triu_indices_from’

[205]

AttributeError: module ‘ndonnx’ has no attribute triu_indices_from

[206]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘true_divide’

[207]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘unique’

[208]

AttributeError: module ‘array_api_compat.dask.array’ has no attribute ‘vander’

[209]

AttributeError: module ‘ndonnx’ has no attribute vander

[210] (1,2)

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘vdot’

[211]

AttributeError: saiunit: backend ‘array_api_compat.dask.array’ has no operation ‘vsplit’

[212]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘vsplit’

[213]

TypeError: transpose() received an invalid combination of arguments - got (Tensor, list), but expected one of: * (Tensor input, int dim0, int dim1) * (Tensor input, name dim0, name dim1)

[214]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘linalg.cholesky’

[215]

AttributeError: saiunit: backend ‘array_api_compat.dask.array’ has no operation ‘linalg.cond’

[216]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘linalg.cond’

[217]

AttributeError: saiunit: backend ‘array_api_compat.dask.array’ has no operation ‘linalg.det’

[218]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘linalg.det’

[219] (1,2)

TypeError: Error interpreting argument to eig as a JAX value. The problematic value is of type <class ‘torch.Tensor’> and was passed to eig at position 0.

[220] (1,2)

TypeError: Error interpreting argument to eig as a JAX value. The problematic value is of type <class ‘dask.array.core.Array’> and was passed to eig at position 0.

[221] (1,2)

TypeError: Error interpreting argument to eig as a JAX value. The problematic value is of type <class ‘ndonnx._array.Array’> and was passed to eig at position 0.

[222] (1,2)

TypeError: Error interpreting argument to transpose as a JAX value. The problematic value is of type <class ‘torch.Tensor’> and was passed to transpose at position 0.

[223] (1,2)

TypeError: Error interpreting argument to transpose as a JAX value. The problematic value is of type <class ‘dask.array.core.Array’> and was passed to transpose at position 0.

[224] (1,2)

TypeError: Error interpreting argument to transpose as a JAX value. The problematic value is of type <class ‘ndonnx._array.Array’> and was passed to transpose at position 0.

[225]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘linalg.inv’

[226]

TypeError: lstsq() got an unexpected keyword argument ‘rcond’

[227] (1,2,3)

AttributeError: module ‘ndonnx’ has no attribute linalg

[228]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘linalg.matrix_norm’

[229]

TypeError: svd() got an unexpected keyword argument ‘compute_uv’

[230]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘linalg.matrix_rank’

[231]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘linalg.matrix_transpose’

[232]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘linalg.norm’

[233]

AttributeError: saiunit: backend ‘array_api_compat.dask.array’ has no operation ‘linalg.pinv’

[234]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘linalg.pinv’

[235]

AttributeError: module ‘array_api_compat.dask.array.linalg’ has no attribute ‘slogdet’

[236]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘linalg.solve’

[237] (1,2)

TypeError: Error interpreting argument to svd as a JAX value. The problematic value is of type <class ‘torch.Tensor’> and was passed to svd at position 0.

[238] (1,2)

TypeError: Error interpreting argument to svd as a JAX value. The problematic value is of type <class ‘dask.array.core.Array’> and was passed to svd at position 0.

[239] (1,2)

TypeError: Error interpreting argument to svd as a JAX value. The problematic value is of type <class ‘ndonnx._array.Array’> and was passed to svd at position 0.

[240]

AttributeError: saiunit: backend ‘array_api_compat.dask.array’ has no operation ‘linalg.tensorinv’

[241]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘linalg.tensorinv’

[242]

AttributeError: saiunit: backend ‘array_api_compat.dask.array’ has no operation ‘linalg.tensorsolve’

[243]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘linalg.tensorsolve’

[244]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘linalg.vector_norm’

[245]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘fft.fft’

[246]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘fft.fft2’

[247] (1,2)

AttributeError: module ‘ndonnx’ has no attribute fft

[248]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘fft.fftn’

[249]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘fft.fftshift’

[250]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘fft.ifft’

[251]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘fft.ifft2’

[252]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘fft.ifftn’

[253]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘fft.ifftshift’

[254]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘fft.irfft’

[255]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘fft.irfft2’

[256]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘fft.irfftn’

[257]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘fft.rfft’

[258]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘fft.rfft2’

[259]

AttributeError: saiunit: backend ‘ndonnx’ has no operation ‘fft.rfftn’

[260] (1,2)

AttributeError: module ‘array_api_compat.torch’ has no attribute ‘copy’

[261] (1,2)

AttributeError: module ‘array_api_compat.dask.array’ has no attribute ‘copy’

[262] (1,2)

AttributeError: module ‘ndonnx’ has no attribute copy

[263] (1,2,3,4,5,6,7)

TypeError: Cannot interpret ‘torch.float64’ as a data type

[264]

InvalidInputException: Argument ‘dask.array<array, shape=(3,), dtype=float64, chunksize=(3,), chunktype=numpy.ndarray>’ of type <class ‘dask.array.core.Array’> is not a valid JAX type.

[265] (1,2,3,4)

TypeError: Cannot interpret ‘Float64’ as a data type

[266]

TypeError: cross() got an unexpected keyword argument ‘axisa’

[267]

AttributeError: module ‘array_api_compat.dask.array’ has no attribute ‘cross’

[268]

AttributeError: module ‘ndonnx’ has no attribute cross

[269] (1,2,3,4,5)

RuntimeError: Unknown backend cuda. Available backends are [‘cpu’]

[270] (1,2,3,4,5)

TypeError: cumprod is not supported for quantities with units (has unit m), because each element of the result would have a different unit exponent. Use .prod() for a single reduction, or convert t…

[271]

TypeError: cumsum() received an invalid combination of arguments - got (Tensor), but expected one of: * (Tensor input, int dim, *, torch.dtype dtype = None, Tensor out = None) * (Tensor input, name…

[272]

TypeError: diagonal() received an invalid combination of arguments - got (Tensor, axis1=int, axis2=int, offset=int), but expected one of: * (Tensor input, *, name outdim, name dim1, name dim2, int …

[273]

AttributeError: module ‘ndonnx’ has no attribute diagonal

[274]

TypeError: Error interpreting argument to <function dot at 0x…> as an abstract array. The problematic value is of type <class ‘ndonnx._array.Array’> and was passed to the function at path a. This…

[275] (1,2,3,4)

ValueError: The dtype of the original data is float32, while we got float64.

[276] (1,2,3,4,5,6,7,8)

BackendError: Quantity.at indexed-update is not supported on the ndonnx backend. Call .to_numpy() (or another concrete backend) on the input first.

[277]

AttributeError: module ‘array_api_compat.dask.array’ has no attribute ‘float16’

[278] (1,2)

AttributeError: ‘Array’ object has no attribute ‘item’

[279] (1,2,3,4,5)

TypeError: nancumprod is not supported for quantities with units (has unit m), because each element of the result would have a different unit exponent. Use .nanprod() for a single reduction, or con…

[280]

TypeError: Error interpreting argument to <function outer at 0x…> as an abstract array. The problematic value is of type <class ‘ndonnx._array.Array’> and was passed to the function at path a. Th…

[281]

NotImplementedError: Quantity.at on dask backend does not support index type ‘Array’ for op ‘set’. Use a boolean mask of the source shape, a slice, an int, or a 1D int array; or call .to_numpy() to…

[282]

ValueError: found array with object dtype but it contains non-string elements

[283] (1,2)

NotImplementedError: Quantity.at on dask backend does not support index type ‘Array’ for op ‘add’. Use a boolean mask of the source shape, a slice, an int, or a 1D int array; or call .to_numpy() to…

[284]

NotImplementedError: Quantity.at on dask backend does not support index type ‘Array’ for op ‘divide’. Use a boolean mask of the source shape, a slice, an int, or a 1D int array; or call .to_numpy()…

[285]

NotImplementedError: Quantity.at on dask backend does not support index type ‘Array’ for op ‘max’. Use a boolean mask of the source shape, a slice, an int, or a 1D int array; or call .to_numpy() to…

[286]

NotImplementedError: Quantity.at on dask backend does not support index type ‘Array’ for op ‘min’. Use a boolean mask of the source shape, a slice, an int, or a 1D int array; or call .to_numpy() to…

[287]

NotImplementedError: Quantity.at on dask backend does not support index type ‘Array’ for op ‘multiply’. Use a boolean mask of the source shape, a slice, an int, or a 1D int array; or call .to_numpy…

[288]

TypeError: split() got an unexpected keyword argument ‘axis’

[289]

AxisError: axis1: axis 0 is out of bounds for array of dimension 0

[290]

TypeError: ‘NoneType’ object is not iterable

[291]

TypeError: int() argument must be a string, a bytes-like object or a real number, not ‘NoneType’

[292]

AttributeError: module ‘ndonnx’ has no attribute swapaxes

[293]

TypeError: Axis value must be an integer, got None

[294]

TypeError: tile(): argument ‘dims’ (position 2) must be tuple of ints, not int

[295]

TypeError: object of type ‘int’ has no len()

[296] (1,2,3,4,5)

cupy backend not installed

[297]

TypeError: Value ‘array(data: [1.0, 2.0, 3.0], dtype=float64)’ with dtype object is not a valid JAX array type. Only arrays of numeric types are supported by JAX.

[298]

ValueError: unable to infer dtype from [1. 2. 3.]

[299]

ValueError: unable to infer dtype from tensor([1., 2., 3.], dtype=torch.float64)

[300]

ValueError: unable to infer dtype from dask.array<array, shape=(3,), dtype=float64, chunksize=(3,), chunktype=numpy.ndarray>

[301]

TypeError: len() of unsized object

[302]

ValueError: ONNX provides no control over the used device

[303]

BackendError (expected): Quantity.tolist() would materialize a dask-backed Quantity. Call q.mantissa.compute() first.

[304]

AttributeError: ‘Array’ object has no attribute ‘tolist’

[305]

TypeError: trace() got an unexpected keyword argument ‘offset’

[306]

AttributeError: module ‘ndonnx’ has no attribute trace

[307]

AttributeError: module ‘ndonnx’ has no attribute transpose

[308]

TypeError: view() received an invalid combination of arguments - got (type), but expected one of: * (torch.dtype dtype) didn’t match because some of the arguments have invalid types: (!type!) * (tu…

[309]

AttributeError: ‘Array’ object has no attribute ‘view’