diff --git a/.github/workflows/main.yml b/.github/workflows/main.yml
index a039b4b..90dce34 100644
--- a/.github/workflows/main.yml
+++ b/.github/workflows/main.yml
@@ -11,6 +11,6 @@ jobs:
     runs-on: ubuntu-latest
     steps:
     - uses: actions/checkout@v2
-    - uses: WebAssembly/wit-abi-up-to-date@v3
+    - uses: WebAssembly/wit-abi-up-to-date@v4
       with:
-        wit-abi-tag: wit-abi-0.1.0
+        wit-abi-tag: wit-abi-0.2.0
diff --git a/wasi-nn.abi.md b/wasi-nn.abi.md
index 1dd8f3f..ab45557 100644
--- a/wasi-nn.abi.md
+++ b/wasi-nn.abi.md
@@ -2,20 +2,15 @@
 
 ## <a href="#tensor_dimensions" name="tensor_dimensions"></a> `tensor-dimensions`: list<`u32`>
 
-  The dimensions of a tensor.
-  
-  The array length matches the tensor rank and each element in the array
-  describes the size of each dimension.
 
 Size: 8, Alignment: 4
 
-## <a href="#tensor_type" name="tensor_type"></a> `tensor-type`: variant
+## <a href="#tensor_type" name="tensor_type"></a> `tensor-type`: enum
 
-  The type of the elements in a tensor.
 
 Size: 1, Alignment: 1
 
-### Variant Cases
+### Enum Cases
 
 - <a href="tensor_type.fp16" name="tensor_type.fp16"></a> [`fp16`](#tensor_type.fp16)
 
@@ -31,19 +26,11 @@ Size: 1, Alignment: 1
 
 ## <a href="#tensor_data" name="tensor_data"></a> `tensor-data`: list<`u8`>
 
-  The tensor data.
-  
-  Initially coneived as a sparse representation, each empty cell would be filled with zeros and
-  the array length must match the product of all of the dimensions and the number of bytes in the
-  type (e.g., a 2x2 tensor with 4-byte f32 elements would have a data array of length 16).
-  Naturally, this representation requires some knowledge of how to lay out data in memory--e.g.,
-  using row-major ordering--and could perhaps be improved.
 
 Size: 8, Alignment: 4
 
 ## <a href="#tensor" name="tensor"></a> `tensor`: record
 
-  A tensor.
 
 Size: 20, Alignment: 4
 
@@ -51,23 +38,15 @@ Size: 20, Alignment: 4
 
 - <a href="tensor.dimensions" name="tensor.dimensions"></a> [`dimensions`](#tensor.dimensions): [`tensor-dimensions`](#tensor_dimensions)
 
-  Describe the size of the tensor (e.g., 2x2x2x2 -> [2, 2, 2, 2]). To represent a tensor
-  containing a single value, use `[1]` for the tensor dimensions.
 
 - <a href="tensor.tensor_type" name="tensor.tensor_type"></a> [`tensor-type`](#tensor.tensor_type): [`tensor-type`](#tensor_type)
 
-  Describe the type of element in the tensor (e.g., f32).
 
 - <a href="tensor.data" name="tensor.data"></a> [`data`](#tensor.data): [`tensor-data`](#tensor_data)
 
-  Contains the tensor data.
 
 ## <a href="#graph_builder" name="graph_builder"></a> `graph-builder`: list<`u8`>
 
-  The graph initialization data.
-  
-  This consists of an array of buffers because implementing backends may encode their graph IR in
-  parts (e.g., OpenVINO stores its IR and weights separately).
 
 Size: 8, Alignment: 4
 
@@ -76,14 +55,12 @@ Size: 8, Alignment: 4
 
 Size: 8, Alignment: 4
 
-## <a href="#graph_encoding" name="graph_encoding"></a> `graph-encoding`: variant
+## <a href="#graph_encoding" name="graph_encoding"></a> `graph-encoding`: enum
 
-  Describes the encoding of the graph. This allows the API to be implemented by various backends
-  that encode (i.e., serialize) their graph IR with different formats.
 
 Size: 1, Alignment: 1
 
-### Variant Cases
+### Enum Cases
 
 - <a href="graph_encoding.openvino" name="graph_encoding.openvino"></a> [`openvino`](#graph_encoding.openvino)
 
@@ -94,13 +71,12 @@ Size: 1, Alignment: 1
 - <a href="graph_encoding.tensorflow" name="graph_encoding.tensorflow"></a> [`tensorflow`](#graph_encoding.tensorflow)
 
 
-## <a href="#execution_target" name="execution_target"></a> `execution-target`: variant
+## <a href="#execution_target" name="execution_target"></a> `execution-target`: enum
 
-  Define where the graph should be executed.
 
 Size: 1, Alignment: 1
 
-### Variant Cases
+### Enum Cases
 
 - <a href="execution_target.cpu" name="execution_target.cpu"></a> [`cpu`](#execution_target.cpu)
 
@@ -111,37 +87,30 @@ Size: 1, Alignment: 1
 - <a href="execution_target.tpu" name="execution_target.tpu"></a> [`tpu`](#execution_target.tpu)
 
 
-## <a href="#error" name="error"></a> `error`: variant
+## <a href="#error" name="error"></a> `error`: enum
 
-  Error codes returned by functions in this API.
 
 Size: 1, Alignment: 1
 
-### Variant Cases
+### Enum Cases
 
 - <a href="error.success" name="error.success"></a> [`success`](#error.success)
 
-  No error occurred.
 
 - <a href="error.invalid_argument" name="error.invalid_argument"></a> [`invalid-argument`](#error.invalid_argument)
 
-  Caller module passed an invalid argument.
 
 - <a href="error.invalid_encoding" name="error.invalid_encoding"></a> [`invalid-encoding`](#error.invalid_encoding)
 
-  Invalid encocing.
 
 - <a href="error.missing_memory" name="error.missing_memory"></a> [`missing-memory`](#error.missing_memory)
 
-  Caller module is missing a memory export.
 
 - <a href="error.busy" name="error.busy"></a> [`busy`](#error.busy)
 
-  Device or resource busy.
 
 - <a href="error.runtime_error" name="error.runtime_error"></a> [`runtime-error`](#error.runtime_error)
 
-  Runtime Error.
 
 # Functions
 
@@ -149,64 +118,59 @@ Size: 1, Alignment: 1
 
 #### <a href="#load" name="load"></a> `load` 
 
-  Load an opaque sequence of bytes to use for inference.
 ##### Params
 
 - <a href="#load.builder" name="load.builder"></a> `builder`: [`graph-builder-array`](#graph_builder_array)
 - <a href="#load.encoding" name="load.encoding"></a> `encoding`: [`graph-encoding`](#graph_encoding)
 - <a href="#load.target" name="load.target"></a> `target`: [`execution-target`](#execution_target)
-##### Results
+##### Result
 
-- <a href="#load." name="load."></a> ``: expected<handle<graph>, [`error`](#error)>
+- expected<handle<graph>, [`error`](#error)>
 
 ----
 
 #### <a href="#init_execution_context" name="init_execution_context"></a> `init-execution-context` 
 
-  Create an execution instance of a loaded graph.
 ##### Params
 
 - <a href="#init_execution_context.graph" name="init_execution_context.graph"></a> `graph`: handle<graph>
-##### Results
+##### Result
 
-- <a href="#init_execution_context." name="init_execution_context."></a> ``: expected<handle<graph-execution-context>, [`error`](#error)>
+- expected<handle<graph-execution-context>, [`error`](#error)>
 
 ----
 
 #### <a href="#set_input" name="set_input"></a> `set-input` 
 
-  Define the inputs to use for inference.
 ##### Params
 
 - <a href="#set_input.ctx" name="set_input.ctx"></a> `ctx`: handle<graph-execution-context>
 - <a href="#set_input.index" name="set_input.index"></a> `index`: `u32`
 - <a href="#set_input.tensor" name="set_input.tensor"></a> `tensor`: [`tensor`](#tensor)
-##### Results
+##### Result
 
-- <a href="#set_input." name="set_input."></a> ``: expected<_, [`error`](#error)>
+- expected<`unit`, [`error`](#error)>
 
 ----
 
 #### <a href="#compute" name="compute"></a> `compute` 
 
-  Compute the inference on the given inputs.
 ##### Params
 
 - <a href="#compute.ctx" name="compute.ctx"></a> `ctx`: handle<graph-execution-context>
-##### Results
+##### Result
 
-- <a href="#compute." name="compute."></a> ``: expected<_, [`error`](#error)>
+- expected<`unit`, [`error`](#error)>
 
 ----
 
 #### <a href="#get_output" name="get_output"></a> `get-output` 
 
-  Extract the outputs after inference.
 ##### Params
 
 - <a href="#get_output.ctx" name="get_output.ctx"></a> `ctx`: handle<graph-execution-context>
 - <a href="#get_output.index" name="get_output.index"></a> `index`: `u32`
-##### Results
+##### Result
 
-- <a href="#get_output." name="get_output."></a> ``: expected<[`tensor`](#tensor), [`error`](#error)>
+- expected<[`tensor`](#tensor), [`error`](#error)>
 
diff --git a/wasi-nn.wit.md b/wasi-nn.wit.md
index 56bceb6..a675026 100644
--- a/wasi-nn.wit.md
+++ b/wasi-nn.wit.md
@@ -107,10 +107,10 @@ index:
 
 ```wit
 // Define the inputs to use for inference.
-set-input: function(ctx: graph-execution-context, index: u32, tensor: tensor) -> expected<_, error>
+set-input: function(ctx: graph-execution-context, index: u32, tensor: tensor) -> expected<unit, error>
 
 // Compute the inference on the given inputs.
-compute: function(ctx: graph-execution-context) -> expected<_, error>
+compute: function(ctx: graph-execution-context) -> expected<unit, error>
 
 // Extract the outputs after inference.
 get-output: function(ctx: graph-execution-context, index: u32) -> expected<tensor, error>