Skip to content

Submission Commands

Register, update, and manage models on the network.

Create a new model (step 1 of create-commit-reveal). Sets up economic parameters (stake, commission) and creates the staking pool. Returns the model ID for subsequent commit and reveal steps.

soma model create [OPTIONS]
ArgumentTypeRequiredDescription
--architecture-versionu64NoArchitecture version (auto-fetched if omitted)
--stake-amountSomaAmountNoAmount of SOMA to stake (auto-fetches minimum if omitted)
--commission-rateu64NoCommission rate in basis points (default 0)
soma model create --commission-rate 500
soma model create --stake-amount 100 --commission-rate 500

Commit model weights (step 2 of create-commit-reveal). Commitments are auto-computed from the weights file, embedding, and decryption key. Works on both created models (initial) and active models (update).

soma model commit <MODEL_ID> [OPTIONS]
ArgumentTypeRequiredDescription
model_idObjectIDYesModel object ID
--weights-filepathYesPath to the encrypted weights file
--weights-urlstringYesURL where the encrypted weights are (or will be) hosted
--embeddingstringYesComma-separated floats for the model embedding
--decryption-keyBase58 (32 bytes)YesBase58-encoded AES-256 decryption key (commitment auto-computed)
soma model commit 0xMODEL... \
--weights-file ./model.bin \
--weights-url https://storage.example.com/weights.bin \
--embedding 0.1,0.2,0.3 \
--decryption-key <base58-key>

Reveal model weights (step 3 of create-commit-reveal). Must be called in the epoch following the commit. Works on both pending models (initial activation) and active models with a pending update.

soma model reveal <MODEL_ID> [OPTIONS]
ArgumentTypeRequiredDescription
model_idObjectIDYesModel object ID
--decryption-keyBase58 (32 bytes)YesBase58-encoded AES-256 decryption key
--embeddingstringYesComma-separated floats for the model embedding
soma model reveal 0xMODEL... \
--decryption-key <base58-key> \
--embedding 0.1,0.2,0.3,0.4

Deactivate a model. Only the model owner can deactivate.

soma model deactivate <MODEL_ID>
ArgumentTypeRequiredDescription
model_idObjectIDYesModel object ID
soma model deactivate 0xMODEL...

Display model information.

soma model info <MODEL_ID>
ArgumentTypeRequiredDescription
model_idObjectIDYesModel object ID
soma model info 0xMODEL...

List all registered models.

soma model list

Set the commission rate for a model.

soma model set-commission-rate <MODEL_ID> --commission-rate <RATE>
ArgumentTypeRequiredDescription
model_idObjectIDYesModel object ID
--commission-rateu16YesCommission rate in basis points (max 10000)
soma model set-commission-rate 0xMODEL... --commission-rate 500

Download model weights via the proxy network.

soma model download <MODEL_ID> [OPTIONS]
ArgumentTypeRequiredDescription
model_idObjectIDYesModel object ID
-o, --outputpathNoOutput file path (default: ./<model_id>.weights)
soma model download 0xMODEL...
soma model download 0xMODEL... -o ./my-model.weights

List targets, submit data, and claim rewards.

List targets with optional filters.

soma target list [OPTIONS]
ArgumentTypeRequiredDescription
-s, --statusstringNoFilter by status: open, filled, claimed
-e, --epochu64NoFilter by epoch
--limitu64NoMax number of results (default 50)
soma target list
soma target list --status open --epoch 5 --limit 20

Show target information.

soma target info <TARGET_ID>
ArgumentTypeRequiredDescription
target_idObjectIDYesTarget object ID
soma target info 0xTARGET...

Submit data to fill a target. Commitment, checksum, and size are auto-computed from the data file. Bond coin is auto-selected.

soma target submit [OPTIONS]
ArgumentTypeRequiredDescription
--target-idObjectIDYesTarget to fill
--data-filepathYesPath to the data file (commitment, checksum, size auto-computed)
--data-urlstringYesURL where data is hosted
--model-idObjectIDYesModel used for the submission (must be in target’s model_ids)
--embeddingstringYesComma-separated floats for the data embedding
--distance-scoref32YesDistance score (must be ≤ target threshold)
--loss-scorestringYesFloat for the loss score from model inference
soma target submit \
--target-id 0xTARGET... \
--data-file ./data.bin \
--data-url https://storage.example.com/data.bin \
--model-id 0xMODEL... \
--embedding 0.1,0.2,0.3 \
--distance-score 0.5 \
--loss-score 0.1

Claim rewards from a filled target. The challenge window must be closed.

soma target claim <TARGET_ID>
ArgumentTypeRequiredDescription
target_idObjectIDYesTarget object ID
soma target claim 0xTARGET...

Download submission data for a target.

soma target download <TARGET_ID> [OPTIONS]
ArgumentTypeRequiredDescription
target_idObjectIDYesTarget object ID
-o, --outputpathNoOutput file path
soma target download 0xTARGET...
soma target download 0xTARGET... -o ./submission-data.bin