Configuration
Configuration options
Configuring with context managers
Methods use_configuration(...) and use_model(...) are designed to work in context managers. Once context manager is left -- old config values are restored.
from inference_sdk import InferenceHTTPClient, InferenceConfiguration
image_url = "https://source.roboflow.com/pwYAXv9BTpqLyFfgQoPZ/u48G0UpWfk8giSw7wrU8/original.jpg"
custom_configuration = InferenceConfiguration(confidence_threshold=0.8)
# Replace ROBOFLOW_API_KEY with your Roboflow API Key
CLIENT = InferenceHTTPClient(
api_url="http://localhost:9001",
api_key="ROBOFLOW_API_KEY"
)
with CLIENT.use_configuration(custom_configuration):
_ = CLIENT.infer(image_url, model_id="soccer-players-5fuqs/1")
with CLIENT.use_model("soccer-players-5fuqs/1"):
_ = CLIENT.infer(image_url)
# after leaving context manager - changes are reverted and `model_id` is still required
_ = CLIENT.infer(image_url, model_id="soccer-players-5fuqs/1")
As you can see -- model_id is required to be given for prediction method only when default model is not configured.
The model id is composed of the string <project_id>/<version_id>. You can find these pieces of information by following the guide on Workspace and Project IDs.
Setting the configuration once and using till next change
Methods configure(...) and select_model(...) are designed to alter the client state and will be preserved until next change.
from inference_sdk import InferenceHTTPClient, InferenceConfiguration
image_url = "https://source.roboflow.com/pwYAXv9BTpqLyFfgQoPZ/u48G0UpWfk8giSw7wrU8/original.jpg"
custom_configuration = InferenceConfiguration(confidence_threshold=0.8)
# Replace ROBOFLOW_API_KEY with your Roboflow API Key
CLIENT = InferenceHTTPClient(
api_url="http://localhost:9001",
api_key="ROBOFLOW_API_KEY"
)
CLIENT.configure(custom_configuration)
CLIENT.infer(image_url, model_id="soccer-players-5fuqs/1")
# custom configuration still holds
CLIENT.select_model(model_id="soccer-players-5fuqs/1")
_ = CLIENT.infer(image_url)
# custom configuration and selected model - still holds
_ = CLIENT.infer(image_url)
One may also initialise in chain mode:
from inference_sdk import InferenceHTTPClient, InferenceConfiguration
# Replace ROBOFLOW_API_KEY with your Roboflow API Key
CLIENT = InferenceHTTPClient(api_url="http://localhost:9001", api_key="ROBOFLOW_API_KEY") \
.select_model("soccer-players-5fuqs/1")
Overriding model_id for specific call
model_id can be overridden for a specific call:
from inference_sdk import InferenceHTTPClient
image_url = "https://source.roboflow.com/pwYAXv9BTpqLyFfgQoPZ/u48G0UpWfk8giSw7wrU8/original.jpg"
# Replace ROBOFLOW_API_KEY with your Roboflow API Key
CLIENT = InferenceHTTPClient(api_url="http://localhost:9001", api_key="ROBOFLOW_API_KEY") \
.select_model("soccer-players-5fuqs/1")
_ = CLIENT.infer(image_url, model_id="another-model/1")
Details about client configuration
inference-client provides InferenceConfiguration dataclass to hold the entire configuration.
from inference_sdk import InferenceConfiguration
Overriding fields in this config changes the behaviour of client (and API serving model). Specific fields are used in specific contexts:
Classification model
visualize_predictions: flag to enable / disable visualisationconfidence_thresholdasconfidencestroke_width: width of stroke in visualisationdisable_preproc_auto_orientation,disable_preproc_contrast,disable_preproc_grayscale,disable_preproc_static_cropto alter server-side pre-processingdisable_active_learningto prevent Active Learning feature from registering the datapointactive_learning_target_dataset-- making inference from specific model (e.g.project_a/1), when you want to save data in another projectproject_b-- the latter should be pointed to by this parametersource-- optional string to set a "source" attribute on the inference call; if using model monitoring, this will get logged with the inference request so you can filter/query inference requests coming from a particular sourcesource_info-- optional string to set additional "source_info" attribute on the inference call
Object detection model
visualize_predictions: flag to enable / disable visualisationvisualize_labels: flag to enable / disable labels visualisation if visualisation is enabledconfidence_thresholdasconfidenceclass_filterto filter out list of classesclass_agnostic_nms: flag to control whether NMS is class-agnosticfix_batch_sizeiou_threshold: to dictate NMS IoU thresholdstroke_width: width of stroke in visualisationmax_detections: max detections to return from modelmax_candidates: max candidates to post-processing from modeldisable_preproc_auto_orientation,disable_preproc_contrast,disable_preproc_grayscale,disable_preproc_static_cropto alter server-side pre-processingdisable_active_learningto prevent Active Learning feature from registering the datapointsource-- optional string for model monitoring source identificationsource_info-- optional string for additional source infoactive_learning_target_dataset-- to save data in a different project
Keypoints detection model
visualize_predictions: flag to enable / disable visualisationvisualize_labels: flag to enable / disable labels visualisation if visualisation is enabledconfidence_thresholdasconfidencekeypoint_confidence_thresholdaskeypoint_confidence-- to filter out detected keypoints based on model confidenceclass_filterto filter out list of object classesclass_agnostic_nms: flag to control whether NMS is class-agnosticfix_batch_sizeiou_threshold: to dictate NMS IoU thresholdstroke_width: width of stroke in visualisationmax_detections: max detections to return from modelmax_candidates: max candidates to post-processing from modeldisable_preproc_auto_orientation,disable_preproc_contrast,disable_preproc_grayscale,disable_preproc_static_cropto alter server-side pre-processingdisable_active_learningto prevent Active Learning feature from registering the datapointsource-- optional string for model monitoring source identificationsource_info-- optional string for additional source infoactive_learning_target_dataset-- to save data in a different project
Instance segmentation model
visualize_predictions: flag to enable / disable visualisationvisualize_labels: flag to enable / disable labels visualisation if visualisation is enabledconfidence_thresholdasconfidenceclass_filterto filter out list of classesclass_agnostic_nms: flag to control whether NMS is class-agnosticfix_batch_sizeiou_threshold: to dictate NMS IoU thresholdstroke_width: width of stroke in visualisationmax_detections: max detections to return from modelmax_candidates: max candidates to post-processing from modeldisable_preproc_auto_orientation,disable_preproc_contrast,disable_preproc_grayscale,disable_preproc_static_cropto alter server-side pre-processingmask_decode_modetradeoff_factordisable_active_learningto prevent Active Learning feature from registering the datapointsource-- optional string for model monitoring source identificationsource_info-- optional string for additional source infoactive_learning_target_dataset-- to save data in a different project
Configuration of client
output_visualisation_format: one of (VisualisationResponseFormat.BASE64,VisualisationResponseFormat.NUMPY,VisualisationResponseFormat.PILLOW) -- given that server-side visualisation is enabled, choose what format should be used in outputimage_extensions_for_directory_scan: while usingCLIENT.infer_on_stream(...)with local directory this parameter controls type of files (extensions) allowed to be processed -- default:["jpg", "jpeg", "JPG", "JPEG", "png", "PNG"]client_downsizing_disabled: set toFalseif you want to perform client-side downsizing -- defaultTrue. Client-side scaling is only supposed to down-scale (keeping aspect-ratio) the input for inference -- to utilise internet connection more efficientlymax_concurrent_requests-- max number of concurrent requests that can be startedmax_batch_size-- max number of elements that can be injected into single requestworkflow_run_retries_enabled-- flag that decides if transient errors in Workflows executions should be retried. Defaults totrueand the default can be altered with environment variable calledWORKFLOW_RUN_RETRIES_ENABLED
Configuration of Workflows execution
profiling_directory: parameter specifying the location where Workflows profiler traces are saved. By default, it is the./inference_profilingdirectory.