|
| 1 | +# CLI Reference |
| 2 | + |
| 3 | +**ep** - eval-protocol: Tools for evaluation and reward modeling |
| 4 | + |
| 5 | +## Global Options |
| 6 | + |
| 7 | +| Option | Type | Default | Required | Description | |
| 8 | +|--------|------|---------|----------|-------------| |
| 9 | +| `--verbose`, `-v` | | false | No | Enable verbose logging | |
| 10 | +| `--profile` | | - | No | Fireworks profile to use (reads ~/.fireworks/profiles/<name>/auth.ini and settings.ini) | |
| 11 | +| `--server` | | - | No | Fireworks API server hostname or URL (e.g., dev.api.fireworks.ai or https://dev.api.fireworks.ai) | |
| 12 | + |
| 13 | +## Commands |
| 14 | + |
| 15 | +### `ep logs` |
| 16 | + |
| 17 | +Serve logs with file watching and real-time updates |
| 18 | + |
| 19 | +| Option | Type | Default | Required | Description | |
| 20 | +|--------|------|---------|----------|-------------| |
| 21 | +| `--port` | int | `8000` | No | Port to bind to (default: 8000) | |
| 22 | +| `--debug` | | false | No | Enable debug mode | |
| 23 | +| `--disable-elasticsearch-setup` | | false | No | Disable Elasticsearch setup | |
| 24 | +| `--use-env-elasticsearch-config` | | false | No | Use env vars for Elasticsearch config (requires ELASTICSEARCH_URL, ELASTICSEARCH_API_KEY, ELASTICSEARCH_INDEX_NAME) | |
| 25 | +| `--use-fireworks` | | false | No | Force Fireworks tracing backend for logs UI (overrides env auto-detection) | |
| 26 | +| `--use-elasticsearch` | | false | No | Force Elasticsearch backend for logs UI (overrides env auto-detection) | |
| 27 | + |
| 28 | +### `ep upload` |
| 29 | + |
| 30 | +Scan for evaluation tests, select, and upload as Fireworks evaluators |
| 31 | + |
| 32 | +| Option | Type | Default | Required | Description | |
| 33 | +|--------|------|---------|----------|-------------| |
| 34 | +| `--path` | | `.` | No | Path to search for evaluation tests (default: current directory) | |
| 35 | +| `--entry` | | - | No | Entrypoint of evaluation test to upload (module:function or path::function). For multiple, separate by commas. | |
| 36 | +| `--id` | | - | No | Evaluator ID to use (if multiple selections, a numeric suffix is appended) | |
| 37 | +| `--display-name` | | - | No | Display name for evaluator (defaults to ID) | |
| 38 | +| `--description` | | - | No | Description for evaluator | |
| 39 | +| `--force` | | false | No | Overwrite existing evaluator with the same ID | |
| 40 | +| `--yes`, `-y` | | false | No | Non-interactive: upload all discovered evaluation tests | |
| 41 | +| `--env-file` | | - | No | Path to .env file containing secrets to upload (default: .env in current directory) | |
| 42 | + |
| 43 | +### `ep create` |
| 44 | + |
| 45 | +Resource creation commands |
| 46 | + |
| 47 | +#### `ep create rft` |
| 48 | + |
| 49 | +Create a Reinforcement Fine-tuning Job on Fireworks |
| 50 | + |
| 51 | +| Option | Type | Default | Required | Description | |
| 52 | +|--------|------|---------|----------|-------------| |
| 53 | +| `--evaluator` | | - | No | Evaluator ID or fully-qualified resource (accounts/{acct}/evaluators/{id}); if omitted, derive from local tests | |
| 54 | +| `--dataset` | | - | No | Use existing dataset (ID or resource 'accounts/{acct}/datasets/{id}') to skip local materialization | |
| 55 | +| `--dataset-jsonl` | | - | No | Path to JSONL to upload as a new Fireworks dataset | |
| 56 | +| `--dataset-builder` | | - | No | Explicit dataset builder spec (module::function or path::function) | |
| 57 | +| `--dataset-display-name` | | - | No | Display name for dataset on Fireworks (defaults to dataset id) | |
| 58 | +| `--base-model` | | - | No | Base model resource id | |
| 59 | +| `--warm-start-from` | | - | No | Addon model to warm start from | |
| 60 | +| `--output-model` | | - | No | Output model id (defaults from evaluator) | |
| 61 | +| `--epochs` | int | `1` | No | - | |
| 62 | +| `--batch-size` | int | `128000` | No | - | |
| 63 | +| `--learning-rate` | float | `3e-05` | No | - | |
| 64 | +| `--max-context-length` | int | `65536` | No | - | |
| 65 | +| `--lora-rank` | int | `16` | No | - | |
| 66 | +| `--gradient-accumulation-steps` | int | - | No | Number of gradient accumulation steps | |
| 67 | +| `--learning-rate-warmup-steps` | int | - | No | Number of LR warmup steps | |
| 68 | +| `--accelerator-count` | int | - | No | - | |
| 69 | +| `--region` | | - | No | Fireworks region enum value | |
| 70 | +| `--display-name` | | - | No | RFT job display name | |
| 71 | +| `--evaluation-dataset` | | - | No | Optional separate eval dataset id | |
| 72 | +| `--eval-auto-carveout` | | true | No | - | |
| 73 | +| `--no-eval-auto-carveout` | | true | No | - | |
| 74 | +| `--chunk-size` | int | `100` | No | Data chunk size for rollout batching | |
| 75 | +| `--temperature` | float | - | No | - | |
| 76 | +| `--top-p` | float | - | No | - | |
| 77 | +| `--top-k` | int | - | No | - | |
| 78 | +| `--max-output-tokens` | int | `32768` | No | - | |
| 79 | +| `--response-candidates-count` | int | `8` | No | - | |
| 80 | +| `--extra-body` | | - | No | JSON string for extra inference params | |
| 81 | +| `--mcp-server` | | - | No | The MCP server resource name to use for the reinforcement fine-tuning job. | |
| 82 | +| `--wandb-enabled` | | false | No | - | |
| 83 | +| `--wandb-project` | | - | No | - | |
| 84 | +| `--wandb-entity` | | - | No | - | |
| 85 | +| `--wandb-run-id` | | - | No | - | |
| 86 | +| `--wandb-api-key` | | - | No | - | |
| 87 | +| `--job-id` | | - | No | Specify an explicit RFT job id | |
| 88 | +| `--yes`, `-y` | | false | No | Non-interactive mode | |
| 89 | +| `--dry-run` | | false | No | Print planned REST calls without sending | |
| 90 | +| `--force` | | false | No | Overwrite existing evaluator with the same ID | |
| 91 | +| `--skip-validation` | | false | No | Skip local dataset and evaluator validation before creating the RFT job | |
| 92 | +| `--ignore-docker` | | false | No | Ignore Dockerfile even if present; run pytest on host during evaluator validation | |
| 93 | +| `--docker-build-extra` | | `` | No | Extra flags to pass to 'docker build' when validating evaluator (quoted string, e.g. "--no-cache --pull --progress=plain") | |
| 94 | +| `--docker-run-extra` | | `` | No | Extra flags to pass to 'docker run' when validating evaluator (quoted string, e.g. "--env-file .env --memory=8g") | |
| 95 | + |
| 96 | +### `ep local-test` |
| 97 | + |
| 98 | +Select an evaluation test and run it locally. If a Dockerfile exists, build and run via Docker; otherwise run on host. |
| 99 | + |
| 100 | +| Option | Type | Default | Required | Description | |
| 101 | +|--------|------|---------|----------|-------------| |
| 102 | +| `--entry` | | - | No | Entrypoint to run (path::function or path). If not provided, a selector will be shown (unless --yes). | |
| 103 | +| `--ignore-docker` | | false | No | Ignore Dockerfile even if present; run pytest on host | |
| 104 | +| `--yes`, `-y` | | false | No | Non-interactive: if multiple tests exist and no --entry, fails with guidance | |
| 105 | +| `--docker-build-extra` | | `` | No | Extra flags to pass to 'docker build' (quoted string, e.g. "--no-cache --pull --progress=plain") | |
| 106 | +| `--docker-run-extra` | | `` | No | Extra flags to pass to 'docker run' (quoted string, e.g. "--env-file .env --memory=8g") | |
0 commit comments