From da2a86417b3b53d9dfe85b6469d20912beca326e Mon Sep 17 00:00:00 2001 From: Andrew Brown Date: Tue, 13 Aug 2024 15:20:49 -0700 Subject: [PATCH] Return errors as strings, not resources This change proposes a different approach to simplifying errors than PR #75 (errors as records). Here we take that approach one step further and completely remove the error code portion entirely, returning only a `string` in the error case. Is this an anti-pattern, though? Some would argue that this is the case, since users might have to resort to string-matching for custom error handling. Others, however, might argue that the error codes were low value anyways since we expect to only see one or maybe two variants as failures to a single call. In any case, this PR offers an alternative to #75 for discussion sake. --- ml.md | 105 +++++------------------------------------------- wit/wasi-nn.wit | 50 +++-------------------- 2 files changed, 16 insertions(+), 139 deletions(-) diff --git a/ml.md b/ml.md index bce656f..2070674 100644 --- a/ml.md +++ b/ml.md @@ -10,7 +10,6 @@ Then, the user passes tensor inputs to the graph, computes the
  • Imports: @@ -90,94 +89,13 @@ containing a single value, use [1] for the tensor dimensions.

    -

    Import interface wasi:nn/errors@0.2.0-rc-2024-06-25

    -

    TODO: create function-specific errors (https://github.com/WebAssembly/wasi-nn/issues/42)

    -
    -

    Types

    -

    enum error-code

    -
    Enum Cases
    - -

    resource error

    -
    -

    Functions

    -

    [constructor]error: func

    -
    Params
    - -
    Return values
    - -

    [method]error.code: func

    -

    Return the error code.

    -
    Params
    - -
    Return values
    - -

    [method]error.data: func

    -

    Errors can propagated with backend specific status through a string value.

    -
    Params
    - -
    Return values
    -

    Import interface wasi:nn/inference@0.2.0-rc-2024-06-25

    An inference "session" is encapsulated by a graph-execution-context. This structure binds a graph to input tensors before compute-ing an inference:

    Types

    -

    type error

    -

    error

    -

    -#### `type tensor` -[`tensor`](#tensor) +

    type tensor

    +

    tensor

    #### `type tensor-data` [`tensor-data`](#tensor_data) @@ -197,7 +115,7 @@ e.g., cannot access a hardware feature requested

    Return values

    [method]graph-execution-context.compute: func

    Compute the inference on the given inputs.

    @@ -210,7 +128,7 @@ https://github.com/WebAssembly/wasi-nn/issues/43.

    Return values

    [method]graph-execution-context.get-output: func

    Extract the outputs after inference.

    @@ -221,18 +139,15 @@ https://github.com/WebAssembly/wasi-nn/issues/43.

    Return values

    Import interface wasi:nn/graph@0.2.0-rc-2024-06-25

    A graph is a loaded instance of a specific ML model (e.g., MobileNet) for a specific ML framework (e.g., TensorFlow):

    Types

    -

    type error

    -

    error

    -

    -#### `type tensor` -[`tensor`](#tensor) +

    type tensor

    +

    tensor

    #### `type graph-execution-context` [`graph-execution-context`](#graph_execution_context) @@ -274,7 +189,7 @@ graph IR in parts (e.g., OpenVINO stores its IR and weights separately).

    Return values

    load: func

    Load a graph from an opaque sequence of bytes to use for inference.

    @@ -286,7 +201,7 @@ graph IR in parts (e.g., OpenVINO stores its IR and weights separately).

    Return values

    load-by-name: func

    Load a graph by name.

    @@ -299,5 +214,5 @@ range from simple to complex (e.g., URLs?) and caching mechanisms of various kin
    Return values
    diff --git a/wit/wasi-nn.wit b/wit/wasi-nn.wit index 872e8cd..b125e64 100644 --- a/wit/wasi-nn.wit +++ b/wit/wasi-nn.wit @@ -12,7 +12,6 @@ world ml { import tensor; import graph; import inference; - import errors; } /// All inputs and outputs to an ML inference are represented as `tensor`s. @@ -61,13 +60,12 @@ interface tensor { /// A `graph` is a loaded instance of a specific ML model (e.g., MobileNet) for a specific ML /// framework (e.g., TensorFlow): interface graph { - use errors.{error}; use tensor.{tensor}; use inference.{graph-execution-context}; /// An execution graph for performing inference (i.e., a model). resource graph { - init-execution-context: func() -> result; + init-execution-context: func() -> result; } /// Describes the encoding of the graph. This allows the API to be implemented by various @@ -96,20 +94,19 @@ interface graph { type graph-builder = list; /// Load a `graph` from an opaque sequence of bytes to use for inference. - load: func(builder: list, encoding: graph-encoding, target: execution-target) -> result; + load: func(builder: list, encoding: graph-encoding, target: execution-target) -> result; /// Load a `graph` by name. /// /// How the host expects the names to be passed and how it stores the graphs for retrieval via /// this function is **implementation-specific**. This allows hosts to choose name schemes that /// range from simple to complex (e.g., URLs?) and caching mechanisms of various kinds. - load-by-name: func(name: string) -> result; + load-by-name: func(name: string) -> result; } /// An inference "session" is encapsulated by a `graph-execution-context`. This structure binds a /// `graph` to input tensors before `compute`-ing an inference: interface inference { - use errors.{error}; use tensor.{tensor, tensor-data}; /// Bind a `graph` to the input and output tensors for an inference. @@ -118,51 +115,16 @@ interface inference { /// (https://github.com/WebAssembly/wasi-nn/issues/43) resource graph-execution-context { /// Define the inputs to use for inference. - set-input: func(name: string, tensor: tensor) -> result<_, error>; + set-input: func(name: string, tensor: tensor) -> result<_, string>; /// Compute the inference on the given inputs. /// /// Note the expected sequence of calls: `set-input`, `compute`, `get-output`. TODO: this /// expectation could be removed as a part of /// https://github.com/WebAssembly/wasi-nn/issues/43. - compute: func() -> result<_, error>; + compute: func() -> result<_, string>; /// Extract the outputs after inference. - get-output: func(name: string) -> result; - } -} - -/// TODO: create function-specific errors (https://github.com/WebAssembly/wasi-nn/issues/42) -interface errors { - enum error-code { - // Caller module passed an invalid argument. - invalid-argument, - // Invalid encoding. - invalid-encoding, - // The operation timed out. - timeout, - // Runtime Error. - runtime-error, - // Unsupported operation. - unsupported-operation, - // Graph is too large. - too-large, - // Graph not found. - not-found, - // The operation is insecure or has insufficient privilege to be performed. - // e.g., cannot access a hardware feature requested - security, - // The operation failed for an unspecified reason. - unknown - } - - resource error { - constructor(code: error-code, data: string); - - /// Return the error code. - code: func() -> error-code; - - /// Errors can propagated with backend specific status through a string value. - data: func() -> string; + get-output: func(name: string) -> result; } }