Make_DSL (neural_nets_lib.Ocannl_tensor.Operation.Make

Parameters

module Grad_spec : sig ... end

Signature

include module type of struct include Grad_spec end

val grad_spec : Tensor.grad_spec

val term : 
  ?init_data:Ir.Assignments.init_data ->
  ?fetch_op:Tensor.fetch_op ->
  Tensor.op_fun

val number : 
  ?label:Base.string Base.list ->
  ?axis_label:Base.string ->
  Base.float ->
  Tensor.t

val bits : 
  ?label:Base.string Base.list ->
  ?axis_label:Base.string ->
  Base.int64 ->
  Tensor.t

val ndarray : Base.float Base.array -> Tensor.op_fun

val threefry4x32 : 
  Tensor.t ->
  Tensor.t ->
  ?label:Base.string Base.list ->
  ?batch_dims:Base.int Base.list ->
  ?batch_axes:(Base.string * Base.int) Base.list ->
  ?input_dims:Base.int Base.list ->
  ?output_dims:Base.int Base.list ->
  ?input_axes:(Base.string * Base.int) Base.list ->
  ?output_axes:(Base.string * Base.int) Base.list ->
  ?deduced:Shape.deduce_within_shape ->
  unit ->
  Tensor.t

val uint4x32_to_prec_uniform : 
  Tensor.t ->
  ?label:Base.string Base.list ->
  ?top_down_prec:Base.bool ->
  ?batch_dims:Base.int Base.list ->
  ?batch_axes:(Base.string * Base.int) Base.list ->
  Tensor.param_op_fun

val uint4x32_to_prec_uniform1 : 
  Tensor.t ->
  ?label:Base.string Base.list ->
  ?top_down_prec:Base.bool ->
  ?batch_dims:Base.int Base.list ->
  ?batch_axes:(Base.string * Base.int) Base.list ->
  Tensor.param_op_fun

val embed_self_id : ?label:Base.string list -> unit -> Tensor.t

val default_param_init : 
  (unit ->
    ?label:Base.string Base.list ->
    ?top_down_prec:Base.bool ->
    ?batch_dims:Base.int Base.list ->
    ?batch_axes:(Base.string * Base.int) Base.list ->
    Tensor.param_op_fun)
    Base.ref

The default initialization operation for param calls.

To avoid user surprises, this defaults to uniform1 which does not impose constraints on the shape of the tensor, but for efficiency, consider setting this to uniform ~grad_spec:Require_grad or normal ~grad_spec:Require_grad instead.

val param : 
  ?value:Base.float ->
  ?values:Base.float Base.array ->
  ?param_init:Tensor.op_fun ->
  Base.string ->
  ?more_label:Base.string Base.list ->
  Tensor.param_op_fun

val einsum : 
  ?capture_dims:Shape.delayed_var_ref Base.list ->
  Base.string ->
  Tensor.t ->
  Tensor.t ->
  Tensor.op_fun

val outer_sum : 
  ?capture_dims:Shape.delayed_var_ref Base.list ->
  Base.string ->
  Tensor.t ->
  Tensor.t ->
  Tensor.op_fun

val einsum1 : 
  ?capture_dims:Shape.delayed_var_ref Base.list ->
  Base.string ->
  Tensor.t ->
  Tensor.op_fun

val einmax1 : 
  ?capture_dims:Shape.delayed_var_ref Base.list ->
  Base.string ->
  Tensor.t ->
  Tensor.op_fun

val tropical : 
  ?capture_dims:Shape.delayed_var_ref Base.list ->
  Base.string ->
  Tensor.t ->
  Tensor.t ->
  Tensor.op_fun

val offsets : Tensor.op_fun

val range : 
  ?label:Base.string list ->
  ?axis_label:Base.string ->
  Base__Int.t ->
  Tensor.t

val range_of_shape : 
  ?label:Base.string list ->
  ?batch_dims:Base.Int.t Base.List.t ->
  ?input_dims:Base.Int.t Base.List.t ->
  ?output_dims:Base.Int.t Base.List.t ->
  ?batch_axes:(Base.string * Base.Int.t) Base.List.t ->
  ?input_axes:(Base.string * Base.Int.t) Base.List.t ->
  ?output_axes:(Base.string * Base.Int.t) Base.List.t ->
  unit ->
  Tensor.t

val stop_gradient : 
  ?spec:Base.string ->
  ?capture_dims:Shape.delayed_var_ref Base.list ->
  Tensor.t ->
  Tensor.op_fun

val reshape : 
  l:Base.string ->
  ?b:Base.int Base.list ->
  ?i:Base.int Base.list ->
  ?o:Base.int Base.list ->
  Ir.Ndarray.t ->
  ?fetch_op:Tensor.fetch_op ->
  ?top_down_prec:Base.bool ->
  ?batch_axes:(Base.string * Base.int) Base.list ->
  ?input_axes:(Base.string * Base.int) Base.list ->
  ?output_axes:(Base.string * Base.int) Base.list ->
  ?deduced:Shape.deduce_within_shape ->
  Base.unit ->
  Tensor.t

val wrap : 
  l:Base.string ->
  ?b:Base.int Base.list ->
  ?i:Base.int Base.list ->
  ?o:Base.int Base.list ->
  Ir.Ndarray.t ->
  ?fetch_op:Tensor.fetch_op ->
  ?top_down_prec:Base.bool ->
  ?batch_axes:(Base.string * Base.int) Base.list ->
  ?input_axes:(Base.string * Base.int) Base.list ->
  ?output_axes:(Base.string * Base.int) Base.list ->
  ?deduced:Shape.deduce_within_shape ->
  Base.unit ->
  Tensor.t

val wrap_padded : 
  l:Base.string ->
  ?b:Base.int Base.list ->
  ?i:Base.int Base.list ->
  ?o:Base.int Base.list ->
  padding:Ir.Ops.axis_padding Base.array ->
  padded_value:Base.float ->
  Ir.Assignments.Nd.t ->
  ?fetch_op:Tensor.fetch_op ->
  ?top_down_prec:Base.bool ->
  ?batch_axes:(Base.string * Base.int) Base.list ->
  ?input_axes:(Base.string * Base.int) Base.list ->
  ?output_axes:(Base.string * Base.int) Base.list ->
  ?deduced:Shape.deduce_within_shape ->
  Base.unit ->
  Tensor.t

val rebatch : 
  l:Base.string ->
  Ir.Ndarray.t ->
  ?fetch_op:Tensor.fetch_op ->
  ?top_down_prec:Base.bool ->
  ?batch_dims:Base.int Base.list ->
  ?batch_axes:(Base.string * Base.int) Base.list ->
  ?deduced:Shape.deduce_within_shape ->
  Base.unit ->
  Tensor.t

val init : 
  l:Base.string ->
  prec:Ir.Ops.prec ->
  ?b:Base.int Base__List.t ->
  ?i:Base.int Base__List.t ->
  ?o:Base.int Base__List.t ->
  f:(Base.int Base.array -> Base.float) ->
  ?fetch_op:Tensor.fetch_op ->
  ?top_down_prec:Base.bool ->
  ?batch_axes:(Base.string * Base.int) Base.list ->
  ?input_axes:(Base.string * Base.int) Base.list ->
  ?output_axes:(Base.string * Base.int) Base.list ->
  ?deduced:Shape.deduce_within_shape ->
  Base.unit ->
  Tensor.t

val uniform : 
  unit ->
  ?label:Base.string Base.list ->
  ?top_down_prec:Base.bool ->
  ?batch_dims:Base.int Base.list ->
  ?batch_axes:(Base.string * Base.int) Base.list ->
  Tensor.param_op_fun

val reshape_param : 
  l:Base.string ->
  ?i:Base.int Base.list ->
  ?o:Base.int Base.list ->
  Ir.Ndarray.t ->
  ?more_label:Base.string Base.list ->
  ?input_axes:(Base.string * Base.int) Base.list ->
  ?output_axes:(Base.string * Base.int) Base.list ->
  ?deduced:Shape.deduce_within_shape ->
  Base.unit ->
  Tensor.t

The input and output dimensions will be inferred if omitted. See reshape.

val wrap_param : 
  l:Base.string ->
  ?i:Base.int Base.list ->
  ?o:Base.int Base.list ->
  Ir.Ndarray.t ->
  ?more_label:Base.string Base.list ->
  ?input_axes:(Base.string * Base.int) Base.list ->
  ?output_axes:(Base.string * Base.int) Base.list ->
  ?deduced:Shape.deduce_within_shape ->
  Base.unit ->
  Tensor.t

See wrap.

val matmul : Tensor.t -> Tensor.t -> Tensor.op_fun

val pointmul : 
  ?spec:Base.string ->
  ?capture_dims:Shape.delayed_var_ref Base.list ->
  Tensor.t ->
  Tensor.t ->
  Tensor.op_fun

val add : 
  ?spec:Base.string ->
  ?capture_dims:Shape.delayed_var_ref Base.list ->
  Tensor.t ->
  Tensor.t ->
  Tensor.op_fun

val pointpow : 
  ?spec:Base.string ->
  ?capture_dims:Shape.delayed_var_ref Base.list ->
  Base.float ->
  Tensor.t ->
  Tensor.op_fun

val relu : 
  ?spec:Base.string ->
  ?capture_dims:Shape.delayed_var_ref Base.list ->
  Tensor.t ->
  Tensor.op_fun

val sat01 : 
  ?spec:Base.string ->
  ?capture_dims:Shape.delayed_var_ref Base.list ->
  Tensor.t ->
  Tensor.op_fun

val fma : Tensor.t -> Tensor.t -> Tensor.t -> Tensor.op_fun

val number_int : 
  ?label:Base.string Base.list ->
  ?axis_label:Base.string ->
  int ->
  Tensor.t

val embed_symbol : 
  ?label:Base.string list ->
  Ir.Indexing.static_symbol ->
  Tensor.t

val embed_dim : ?label:Base.string list -> Shape.delayed_var_ref -> Tensor.t

val sub : 
  ?spec:Base.string ->
  ?capture_dims:Shape.delayed_var_ref Base.list ->
  Tensor.t ->
  Tensor.t ->
  Tensor.op_fun

val pointdiv : 
  ?spec:Base.string ->
  ?capture_dims:Shape.delayed_var_ref Base.list ->
  Tensor.t ->
  Tensor.t ->
  Tensor.op_fun

val slice : Idx.static_symbol -> Tensor.t -> Tensor.op_fun

val exp : 
  ?spec:Base.string ->
  ?capture_dims:Shape.delayed_var_ref Base.list ->
  Tensor.t ->
  Tensor.op_fun

val log : 
  ?spec:Base.string ->
  ?capture_dims:Shape.delayed_var_ref Base.list ->
  Tensor.t ->
  Tensor.op_fun

val log2 : 
  ?spec:Base.string ->
  ?capture_dims:Shape.delayed_var_ref Base.list ->
  Tensor.t ->
  Tensor.op_fun

val sin : 
  ?spec:Base.string ->
  ?capture_dims:Shape.delayed_var_ref Base.list ->
  Tensor.t ->
  Tensor.op_fun

val cos : 
  ?spec:Base.string ->
  ?capture_dims:Shape.delayed_var_ref Base.list ->
  Tensor.t ->
  Tensor.op_fun

val neg : 
  ?spec:Base.string ->
  ?capture_dims:Shape.delayed_var_ref Base.list ->
  Tensor.t ->
  Tensor.op_fun

val sqrt : 
  ?spec:Base.string ->
  ?capture_dims:Shape.delayed_var_ref Base.list ->
  Tensor.t ->
  Tensor.op_fun

val recip : 
  ?spec:Base.string ->
  ?capture_dims:Shape.delayed_var_ref Base.list ->
  Tensor.t ->
  Tensor.op_fun

val recip_sqrt : 
  ?spec:Base.string ->
  ?capture_dims:Shape.delayed_var_ref Base.list ->
  Tensor.t ->
  Tensor.op_fun

val tanh : 
  ?spec:Base.string ->
  ?capture_dims:Shape.delayed_var_ref Base.list ->
  Tensor.t ->
  Tensor.op_fun

val where : Tensor.t -> Tensor.t -> Tensor.t -> Tensor.op_fun

val not : 
  ?spec:Base.string ->
  ?capture_dims:Shape.delayed_var_ref Base.list ->
  Tensor.t ->
  Tensor.op_fun

val lt : 
  ?spec:Base.string ->
  ?capture_dims:Shape.delayed_var_ref Base.list ->
  Tensor.t ->
  Tensor.t ->
  Tensor.op_fun

val eq : 
  ?spec:Base.string ->
  ?capture_dims:Shape.delayed_var_ref Base.list ->
  Tensor.t ->
  Tensor.t ->
  Tensor.op_fun

val ne : 
  ?spec:Base.string ->
  ?capture_dims:Shape.delayed_var_ref Base.list ->
  Tensor.t ->
  Tensor.t ->
  Tensor.op_fun

val uniform_at : 
  Tensor.t ->
  ?label:Base.string Base.list ->
  ?top_down_prec:Base.bool ->
  ?batch_dims:Base.int Base.list ->
  ?batch_axes:(Base.string * Base.int) Base.list ->
  Tensor.param_op_fun

val uniform1 : 
  unit ->
  ?label:Base.string Base.list ->
  ?top_down_prec:Base.bool ->
  ?batch_dims:Base.int Base.list ->
  ?batch_axes:(Base.string * Base.int) Base.list ->
  Tensor.param_op_fun

val uniform_at1 : 
  Tensor.t ->
  ?label:Base.string Base.list ->
  ?top_down_prec:Base.bool ->
  ?batch_dims:Base.int Base.list ->
  ?batch_axes:(Base.string * Base.int) Base.list ->
  Tensor.param_op_fun

val interleave : Tensor.t -> Tensor.t -> Tensor.op_fun

module O : sig ... end