- Argamak now supports
gleam_stdlib
v1.0. - Argamak now requires Gleam v0.34 or later.
- Argamak now requires Gleam v0.33 or later.
- The
tensor
module'sconcat
function now returns anAxisNotFound
error when the given find function isFalse
for everyAxis
.
- Argamak now requires Gleam v0.30 or later.
- The
tensor
module gains theconcat
slicing/joining function.
- The
tensor
module gains theTensorResult
type; thefrom_bool
andfrom_bools
creation functions; thesize
reflection function; thesqueeze
transformation function; theequal
,not_equal
,greater
,greater_or_equal
,less
,less_or_equal
,logical_and
,logical_or
,logical_xor
, andlogical_not
logical functions; theadd
,subtract
,multiply
,divide
,try_divide
,remainder
,try_remainder
,modulo
,try_modulo
,power
,max
, andmin
arithmetic functions; theabsolute_value
,negate
,sign
,ceiling
,floor
,round
,exp
,square_root
, andln
basic math functions; theall
,in_situ_all
,any
,in_situ_any
,arg_max
,in_situ_arg_max
,arg_min
,in_situ_arg_min
,max_over
,in_situ_max_over
,min_over
,in_situ_min_over
,sum
,in_situ_sum
,product
,in_situ_product
,mean
, andin_situ_mean
reduction functions; theto_bool
,to_floats
,to_ints
, andto_bools
conversion functions; and thedebug
andprint_data
utility functions. - The
tensor
module'sas_format
function has been renamed toreformat
and now takes aFormat
record instead of a function reference. - The
tensor
module no longer includes theto_list
function. - The
Tensor
type signature now includes only the numeric format as a generic. - The
axis
module has been added with theAxis
andAxes
types; thename
andsize
reflection functions; and therename
andresize
transformation functions. - The
space
module gains thefrom_list
creation function; and themap
andmerge
transformation functions. - The
space
module no longer includes theelements
andmap_elements
functions. - The
space
module'sd0
function has been renamed tonew
and now returns an emptySpace
record directly. - The
space
module and itsSpace
andSpaceError
types have been reworked: TheSpace
type signature no longer includes any generics, and the constructorsD0
throughD6
have been removed. - The
Format
type has been reworked and now includes the numeric format as a generic. - Several numeric format types have been added to the
format
module. - The
util
module has been removed.
- The
space
module gets an updatedd1
function so the dimension size can be given. - The
tensor
module gains thebroadcast
, andbroadcast_over
functions for all compilation targets. - Argamak now compiles and runs with the JavaScript target.
- Argamak now uses the
gleam
build tool.
- Initial release!
- The
format
module gains theFormat
andNative
(JavaScript planned) types for all compilation targets, along with thefloat32
,int32
,to_native
, andto_string
functions for all compilation targets (JavaScript planned) and thebfloat16
,float16
,float64
,int16
,int64
,int8
,uint16
,uint32
,uint64
, anduint8
functions for the Erlang compilation target. - The
space
module gains theD0
,D1
,D2
,D3
,D4
,D5
,D6
,Space
, andSpaceError
(withSpaceErrors
alias) types for all compilation targets, along with theaxes
,d0
,d1
,d2
,d3
,d4
,d5
,d6
,degree
,elements
,map_elements
,shape
, andto_string
functions for all compilation targets (JavaScript planned). - The
tensor
module gains theNative
(JavaScript planned),Tensor
, andTensorError
types for all compilation targets, along with theas_format
,axes
,format
,from_float
,from_floats
,from_int
,from_ints
,from_native
,print
,rank
,reshape
,shape
,space
,to_float
,to_int
,to_list
, andto_native
functions for all compilation targets (JavaScript planned). - The
util
module gains theUtilError
type for all compilation targets, along with therecord_to_string
function for all compilation targets.