Model management

The cfsutil ai model commands let you manage AI models within a workspace. Use these commands to add new models, list configured models, update model properties, and remove models from a workspace.

Add command

Adds a model to a workspace. The command adds a new entry to the AIModels array in the .cfsconfig file. For example, adding a TFLM model targeting CM4 produces an entry like:

{
  "Name": "hello_world_f32",
  "Files": {
    "Model": "path/to/hello_world_f32.tflite"
  },
  "Target": {
    "Core": "CM4"
  },
  "Backend": {
    "Name": "tflm",
    "Extensions": {}
  },
  "Enabled": true
}

The model name is derived from the file name.

Note

A model with the same name cannot be added twice. The name is taken from the model file name (without extension) and must not exceed 64 characters. If the derived name is longer, rename the model file to a shorter name.

cfsutil ai model add --core <value> -m <value> [--format text|json] [-w <value>] [--config <value>] [--dataset <value>] [-a <value>] [--network-config <value>] [-e <value>...] [--ignore-cache]

Required flags

Flag	Short	Description
--model	-m	Path or URL to the model file.
--config or --workspace		One must be provided. --config takes precedence over --workspace if both are given.
--core		Target core (for example, CM4, CM33).

Example

cfsutil ai model add --config .cfs/max32690.cfsconfig --core CM4 --model path/to/model.tflite

Sample output

Model added successfully.

C/C++ build compatibility

For TFLM models, cfsutil ai build generates C++ source and header files (.cpp, .hpp). If your project uses a non-AI template with main.c, you may see compiler warnings when integrating the generated files. Replace main.c with main.cpp or use a C++ wrapper.

Optional flags

Flag	Short	Description
--acc	-a	Target accelerator (for example, CNN).
--extension	-e	Backend-specific fields as key=value pairs. Can be repeated. For example, for tflm: -e Symbol=my_model -e Section=.data; for izer: -e Softmax=true -e ClockDivider=4. Use cfsutil ai backends list --name <backend> to see available extension fields for a given backend.
--dataset		Path or URL to a sample dataset for quantization calibration (for example, a folder of representative input images).
--ignore-cache		Bypass the cache and fetch the latest remote files.
--network-config		Path or URL to the Izer network configuration YAML. Required for izer backends.
--format		Output format: text (default) or json.

Note

For reference on how to train PyTorch models compatible with MAX78002, see the open-source ai8x-synthesis repository.

Add a TFLM model with custom symbol and memory section

cfsutil ai model add \
    --config .cfs/max32690.cfsconfig \
    --core CM4 \
    --model path/to/hello_world_f32.tflite \
    -e Symbol=hello_world_model \
    -e Section=.data

Add an izer model for MAX78002 with network config

--acc CNN selects the izer backend, which activates --network-config. Both --model and --network-config accept local paths or URLs.

cfsutil ai model add \
    --config .cfs/max78002.cfsconfig \
    --core CM4 \
    --acc CNN \
    --model path/to/model.pth.tar \
    --network-config path/to/network.yaml \
    -e Softmax=true

Warning

If --acc CNN is omitted, the model is treated as a tflm model and --network-config is silently ignored.

List command

Lists the AI models configured in a workspace. Use --core to filter by a specific core, or --verbose to show full model details.

cfsutil ai model list [--format text|json] [-w <value>] [--config <value>] [-c <value>] [--verbose]

Required flags

Flag	Short	Description
--config or --workspace		One must be provided. --config takes precedence over --workspace if both are given.

Optional flags

Flag	Short	Description
--core	-c	Filter models by target core.
--format		Output format: text (default) or json.
--verbose		Show file name, backend, and extension details for each model.

Example

cfsutil ai model list --config .cfs/max32690-tqfn.cfsconfig
cfsutil ai model list --config .cfs/max32690-tqfn.cfsconfig --core CM4 --verbose

Sample output (verbose)

CM4:
  Hello World Float 32
    Backend: tflm
    Files:
      Model: ./m4/hello_world_f32.tflite
    Extensions:
      Symbol: hello_world_model_f32
      Section: .data
    Enabled: true

Remove command

Removes a model from a workspace by name. The command deletes the corresponding entry from the AIModels array in the .cfsconfig file.

cfsutil ai model remove --name <value> [--format text|json] [-w <value>] [--config <value>]

Required flags

Flag	Short	Description
--name		Name of the model to remove. Use quotes if the name contains spaces, for example --name "Hello World Float 32".
--config or --workspace		One must be provided. --config takes precedence over --workspace if both are given.

Optional flags

Flag	Short	Description
--format		Output format: text (default) or json.

Example

cfsutil ai model remove --name hello_world_f32 --config .cfs/max32690-tqfn.cfsconfig

Update command

Updates properties of an existing model in a workspace. Use --set key=value to specify the fields to change. The --set flag can be repeated to update multiple fields in one command.

cfsutil ai model update --name <value> --set <value>... [--format text|json] [-w <value>] [--config <value>] [-c <value>] [--ignore-cache]

Required flags

Flag	Short	Description
--name		Name of the model to update. Use quotes if the name contains spaces, for example --name "Hello World Float 32".
--set		Field to update as key=value. Can be repeated.
--config or --workspace		One must be provided. --config takes precedence over --workspace if both are given.

Optional flags

Flag	Short	Description
--core	-c	Target core.
--ignore-cache		Bypass the cache and fetch the latest remote files.
--format		Output format: text (default) or json.

Updatable fields

Core fields

Key	Description
enabled	Enable or disable the model (true or false).
model	Replace the model file path or URL.
name	Rename the model. Must not conflict with an existing model name and must not exceed 64 characters.

Update core fields

cfsutil ai model update --config .cfs/max32690-tqfn.cfsconfig --name hello_world_f32 --set name=hello_world_renamed
cfsutil ai model update --config .cfs/max32690-tqfn.cfsconfig --name hello_world_f32 --set model=path/to/new_model.tflite
cfsutil ai model update --config .cfs/max32690-tqfn.cfsconfig --name hello_world_f32 --set enabled=false

tflm extension fields

Key	Description
ArenaSection	Memory section for the arena.
ArenaSize	Arena size for the model.
Dataset	Dataset file.
DatasetSection	Memory section for the dataset.
Section	Memory section for model data.
Symbol	Symbol name for model data.

Update tflm extension fields

cfsutil ai model update --config .cfs/max32690-tqfn.cfsconfig --name hello_world_f32 --set Symbol=hello_world_model --set Section=.data

izer extension fields

Key	Description
AvgPoolRounding	Round average pooling results.
ClockDivider	CNN clock divider.
Fifo	Use FIFO.
InputShape	Input shape.
Prefix	Test name prefix.
NetworkConfig	Izer network configuration path or URL.
Softmax	Enable softmax layer generation.
Timer	Inference timer.

Update izer extension fields

cfsutil ai model update --config .cfs/max78002-csbga.cfsconfig --name cats-dogs --set Softmax=true --set ClockDivider=4

Tip

Use cfsutil ai backends list --name <backend> to see all available extension fields for a given backend.