--output_type

Applicability

Product	Supported (Yes/No)
Atlas 350 Accelerator Card	√
Atlas A3 training product/Atlas A3 inference product	√
Atlas A2 training product/Atlas A2 inference product	√
Atlas 200I/500 A2 inference product	√
Atlas inference product	√
Atlas training product	√

Description

Sets the output data type of a network or an output node.

Argument

Argument:

FP32: recommended for classification and object detection networks
FP16: recommended for classification and object detection networks. It is usually used when the output of one network is used as the input of another.
UINT8: recommended for image super-resolution networks for better inference performance.
INT8
UINT16
INT16
UINT32
INT32
UINT64
INT64
DOUBLE
HIF8: This type is only supported by Atlas 350 Accelerator Card.
FP8E5M2: This type is only supported by Atlas 350 Accelerator Card.
FP8E4M3FN: This type is only supported by Atlas 350 Accelerator Card.

Restrictions:

After model conversion, the preceding data types are represented in the corresponding offline model (.om) file as follows:

DT_FLOAT
DT_FLOAT16
DT_UINT8
DT_INT8
DT_UINT16
DT_INT16
DT_UINT32
DT_INT32
DT_UINT64
DT_INT64
DT_DOUBLE
DT_HIFLOAT8
DT_FLOAT8_E5M2
DT_FLOAT8_E4M3FN

If the output data type of the network is not specified during model conversion, the data type of the operator output at the output layer of the original network model is used. If the type is specified, the type specified by this option is used and the type specified by --is_output_adjust_hw_layout does not take effect.

Suggestions and Benefits

None

Example

Output data type of a network:
```
--output_type=FP32
```
Output data type of an output node:
For example, --output_type="node1:0:FP16;node2:0:FP32" indicates that the output data type of node1 is set to FP16 and that of node2 is set to FP32. Enclose all the nodes in double quotation marks (""), and separate the nodes with semicolons (;).
In this scenario, use this option in conjunction with --out_nodes.
```
--model=$HOME/module/resnet50_tensorflow.pb --framework=3 --output=$HOME/module/out/tf_resnet50  --soc_version=<soc_version>  --output_type="conv1:0:FP16"  --out_nodes="conv1:0"
```

Dependencies and Restrictions