1- package wasi : nn
1+ package wasi : nn ;
22
33/// `wasi-nn` is a WASI API for performing machine learning (ML) inference. The API is not (yet)
44/// capable of performing ML training. WebAssembly programs that want to use a host's ML
@@ -9,10 +9,10 @@ package wasi:nn
99///
1010/// This example world shows how to use these primitives together.
1111world ml {
12- import tensor
13- import graph
14- import inference
15- import errors
12+ import tensor ;
13+ import graph ;
14+ import inference ;
15+ import errors ;
1616}
1717
1818/// All inputs and outputs to an ML inference are represented as `tensor` s.
@@ -21,7 +21,7 @@ interface tensor {
2121 ///
2222 /// The array length matches the tensor rank and each element in the array describes the size of
2323 /// each dimension
24- type tensor-dimensions = list <u32 >
24+ type tensor-dimensions = list <u32 >;
2525
2626 /// The type of the elements in a tensor.
2727 enum tensor-type {
@@ -41,7 +41,7 @@ interface tensor {
4141 /// in the type (e.g., a 2x2 tensor with 4-byte f32 elements would have a data array of length
4242 /// 16). Naturally, this representation requires some knowledge of how to lay out data in
4343 /// memory--e.g., using row-major ordering--and could perhaps be improved.
44- type tensor-data = list <u8 >
44+ type tensor-data = list <u8 >;
4545
4646 record tensor {
4747 // Describe the size of the tensor (e.g., 2x2x2x2 -> [2, 2, 2, 2]). To represent a tensor
@@ -59,13 +59,13 @@ interface tensor {
5959/// A `graph` is a loaded instance of a specific ML model (e.g., MobileNet) for a specific ML
6060/// framework (e.g., TensorFlow):
6161interface graph {
62- use errors . {error }
63- use tensor . {tensor }
62+ use errors . {error };
63+ use tensor . {tensor };
6464
6565 /// An execution graph for performing inference (i.e., a model).
6666 ///
6767 /// TODO: replace with `resource` (https://github.com/WebAssembly/wasi-nn/issues/47).
68- type graph = u32
68+ type graph = u32 ;
6969
7070 /// Describes the encoding of the graph. This allows the API to be implemented by various
7171 /// backends that encode (i.e., serialize) their graph IR with different formats.
@@ -89,45 +89,45 @@ interface graph {
8989 ///
9090 /// This gets bundled up into an array of buffers because implementing backends may encode their
9191 /// graph IR in parts (e.g., OpenVINO stores its IR and weights separately).
92- type graph-builder = list <u8 >
92+ type graph-builder = list <u8 >;
9393
9494 /// Load a `graph` from an opaque sequence of bytes to use for inference.
95- load : func (builder : list <graph-builder >, encoding : graph-encoding , target : execution-target ) -> result <graph , error >
95+ load : func (builder : list <graph-builder >, encoding : graph-encoding , target : execution-target ) -> result <graph , error >;
9696
9797 /// Load a `graph` by name.
9898 ///
9999 /// How the host expects the names to be passed and how it stores the graphs for retrieval via
100100 /// this function is **implementation-specific** . This allows hosts to choose name schemes that
101101 /// range from simple to complex (e.g., URLs?) and caching mechanisms of various kinds.
102- load-by-name : func (name : string ) -> result <graph , error >
102+ load-by-name : func (name : string ) -> result <graph , error >;
103103}
104104
105105/// An inference "session" is encapsulated by a `graph-execution-context` . This structure binds a
106106/// `graph` to input tensors before `compute` -ing an inference:
107107interface inference {
108- use errors . {error }
109- use tensor . {tensor , tensor-data }
110- use graph . {graph }
108+ use errors . {error };
109+ use tensor . {tensor , tensor-data };
110+ use graph . {graph };
111111
112112 /// Bind a `graph` to the input and output tensors for an inference.
113113 ///
114114 /// TODO: this is no longer necessary in WIT (https://github.com/WebAssembly/wasi-nn/issues/43)
115- type graph-execution-context = u32
115+ type graph-execution-context = u32 ;
116116
117117 /// Create an execution instance of a loaded graph.
118- init-execution-context : func (graph : graph ) -> result <graph-execution-context , error >
118+ init-execution-context : func (graph : graph ) -> result <graph-execution-context , error >;
119119
120120 /// Define the inputs to use for inference.
121- set-input : func (ctx : graph-execution-context , index : u32 , tensor : tensor ) -> result <_ , error >
121+ set-input : func (ctx : graph-execution-context , index : u32 , tensor : tensor ) -> result <_ , error >;
122122
123123 /// Compute the inference on the given inputs.
124124 ///
125125 /// Note the expected sequence of calls: `set-input` , `compute` , `get-output` . TODO: this
126126 /// expectation could be removed as a part of https://github.com/WebAssembly/wasi-nn/issues/43.
127- compute : func (ctx : graph-execution-context ) -> result <_ , error >
127+ compute : func (ctx : graph-execution-context ) -> result <_ , error >;
128128
129129 /// Extract the outputs after inference.
130- get-output : func (ctx : graph-execution-context , index : u32 ) -> result <tensor-data , error >
130+ get-output : func (ctx : graph-execution-context , index : u32 ) -> result <tensor-data , error >;
131131}
132132
133133/// TODO: create function-specific errors (https://github.com/WebAssembly/wasi-nn/issues/42)
0 commit comments