Skip to content

Environment Variables Configuration

This guide covers all environment variables available in inference-models for configuring model loading, caching, API access, and runtime behavior.

Quick Start

Set environment variables before importing inference-models:

# Set API key
export ROBOFLOW_API_KEY="your_api_key_here"

# Set model cache directory
export INFERENCE_HOME="/path/to/cache"

# Set device
export DEFAULT_DEVICE="cuda:0"

inference-models reads the process environment when it is first imported and then loads a .env file from the current working directory without overriding already-set process values. This matches the inference package, so package import order cannot change the selected cache or offline mode.

Core Configuration

API Authentication

ROBOFLOW_API_KEY (or API_KEY)
Your Roboflow API key for accessing models.

export ROBOFLOW_API_KEY="your_api_key_here"

Get your API key from: https://docs.roboflow.com/api-reference/authentication

ROBOFLOW_ENVIRONMENT
Environment to use: prod (default) or staging.

export ROBOFLOW_ENVIRONMENT="prod"

ROBOFLOW_API_HOST
Override API host URL (auto-set based on environment).

export ROBOFLOW_API_HOST="https://api.roboflow.com"

Model Cache

INFERENCE_HOME
Directory where downloaded models are cached. If it is unset, MODEL_CACHE_DIR is used; if both are unset, the default is /tmp/cache.

export INFERENCE_HOME="/home/user/.cache/inference-models"

OFFLINE_MODE Startup-only switch for loading network-provider models exclusively from a trusted, compatible local cache.

export OFFLINE_MODE="True"

The first import of either inference or inference_models latches this value for the process. Changing or removing the variable later does not change the mode; restart the process instead. Child processes inherit the latch when they inherit the parent environment with Inference's private marker intact. That marker is trusted internal process state, not a security boundary against arbitrary code already running in the process. A child launched with a deliberately rewritten or sanitized environment is a new startup boundary, so use operating-system or network-level isolation when a hard air gap is required.

The startup latch also enables the Hugging Face and Ultralytics dependency offline controls before Inference imports those libraries, preventing their built-in connectivity checks and online-gated behavior. If HF_HOME is not explicitly configured, both packages set it before heavy imports to $INFERENCE_HOME/hf_home, $MODEL_CACHE_DIR/hf_home, or /tmp/cache/hf_home in that order. This keeps implicit Hugging Face backbone, processor, and checkpoint downloads in the mounted cache across the online-warm and fresh-offline phases.

Warm the cache online with the matching inference-models release and the same model-loading constraints and runtime environment before enabling offline mode. Legacy cache manifests do not contain the canonical owner, trust, dependency, and compatibility metadata required by the offline loader and must be re-warmed. A credential-free offline restart can use a cache warmed with a key only when the current metadata proves one unambiguous canonical model identity. A changed or rotated non-empty key requires an exact matching cache entry and otherwise fails closed.

Device Selection

DEFAULT_DEVICE
Default device for model inference: cpu, cuda, cuda:0, etc.

export DEFAULT_DEVICE="cuda:0"  # Use first GPU
export DEFAULT_DEVICE="cpu"     # Use CPU

API Configuration

Request Settings

API_CALLS_TIMEOUT
Timeout for API calls in seconds. Default: 5

export API_CALLS_TIMEOUT="10"

API_CALLS_MAX_TRIES
Maximum retry attempts for API calls. Default: 3

export API_CALLS_MAX_TRIES="5"

IDEMPOTENT_API_REQUEST_CODES_TO_RETRY
HTTP status codes to retry (comma-separated). Default: 408,429,502,503,504

export IDEMPOTENT_API_REQUEST_CODES_TO_RETRY="408,429,500,502,503,504"

Backend Configuration

ONNX Runtime

ONNXRUNTIME_EXECUTION_PROVIDERS Override ONNX execution providers, comma separated, no spaces. Default: CUDAExecutionProvider,OpenVINOExecutionProvider,CoreMLExecutionProvider,CPUExecutionProvider

export ONNX_EXECUTION_PROVIDERS="CPUExecutionProvider"

Prediction Parameter Defaults

These environment variables control the default values for prediction parameters across all models. Individual models may override these defaults. See Prediction Parameters for detailed information about each parameter.

General Detection Parameters

INFERENCE_MODELS_DEFAULT_CONFIDENCE Default confidence threshold for filtering predictions. Default: 0.4

export INFERENCE_MODELS_DEFAULT_CONFIDENCE="0.5"

INFERENCE_MODELS_DEFAULT_IOU_THRESHOLD Default IoU threshold for Non-Maximum Suppression. Default: 0.3

export INFERENCE_MODELS_DEFAULT_IOU_THRESHOLD="0.45"

INFERENCE_MODELS_DEFAULT_MAX_DETECTIONS Default maximum number of detections to return. Default: 300

export INFERENCE_MODELS_DEFAULT_MAX_DETECTIONS="100"

INFERENCE_MODELS_INSTANCE_SEG_MASK_PROCESSING_CHUNK_SIZE Number of instance-segmentation masks upscaled to original resolution per slice when producing dense masks. Bounds postprocessing memory (the float32 working set is chunk x H x W instead of detections x H x W). Must be >= 1. Default: 16

export INFERENCE_MODELS_INSTANCE_SEG_MASK_PROCESSING_CHUNK_SIZE="16"

INFERENCE_MODELS_DEFAULT_CLASS_AGNOSTIC_NMS Default for class-agnostic NMS. Default: false

export INFERENCE_MODELS_DEFAULT_CLASS_AGNOSTIC_NMS="true"

General Vision-Language Model Parameters

INFERENCE_MODELS_DEFAULT_MAX_NEW_TOKENS Default maximum number of tokens to generate. Default: 4096

export INFERENCE_MODELS_DEFAULT_MAX_NEW_TOKENS="1000"

INFERENCE_MODELS_DEFAULT_NUM_BEAMS Default number of beams for beam search. Default: 3

export INFERENCE_MODELS_DEFAULT_NUM_BEAMS="5"

INFERENCE_MODELS_DEFAULT_DO_SAMPLE Default for sampling during generation. Default: false

export INFERENCE_MODELS_DEFAULT_DO_SAMPLE="true"

INFERENCE_MODELS_DEFAULT_SKIP_SPECIAL_TOKENS Default for skipping special tokens in output. Default: false

export INFERENCE_MODELS_DEFAULT_SKIP_SPECIAL_TOKENS="true"

Model-Specific Overrides

Individual models can override the general defaults. Below are model-specific environment variables:

DeepLabV3+

INFERENCE_MODELS_DEEP_LAB_V3_PLUS_DEFAULT_CONFIDENCE Default: 0.5

export INFERENCE_MODELS_DEEP_LAB_V3_PLUS_DEFAULT_CONFIDENCE="0.6"

DINOv3

INFERENCE_MODELS_DINOV3_DEFAULT_CONFIDENCE Default: 0.5

export INFERENCE_MODELS_DINOV3_DEFAULT_CONFIDENCE="0.6"

EasyOCR

INFERENCE_MODELS_EASYOCR_DEFAULT_CONFIDENCE Default: 0.3

export INFERENCE_MODELS_EASYOCR_DEFAULT_CONFIDENCE="0.4"

Florence-2

INFERENCE_MODELS_FLORENCE2_DEFAULT_MAX_NEW_TOKENS Default: Inherits from INFERENCE_MODELS_DEFAULT_MAX_NEW_TOKENS

export INFERENCE_MODELS_FLORENCE2_DEFAULT_MAX_NEW_TOKENS="2048"

INFERENCE_MODELS_FLORENCE2_DEFAULT_NUM_BEAMS Default: Inherits from INFERENCE_MODELS_DEFAULT_NUM_BEAMS

export INFERENCE_MODELS_FLORENCE2_DEFAULT_NUM_BEAMS="5"

INFERENCE_MODELS_FLORENCE2_DEFAULT_DO_SAMPLE Default: Inherits from INFERENCE_MODELS_DEFAULT_DO_SAMPLE

export INFERENCE_MODELS_FLORENCE2_DEFAULT_DO_SAMPLE="true"

Grounding DINO

INFERENCE_MODELS_GROUNDING_DINO_DEFAULT_BOX_CONFIDENCE Default: 0.5

export INFERENCE_MODELS_GROUNDING_DINO_DEFAULT_BOX_CONFIDENCE="0.6"

INFERENCE_MODELS_GROUNDING_DINO_DEFAULT_MAX_DETECTIONS Default: Inherits from INFERENCE_MODELS_DEFAULT_MAX_DETECTIONS

export INFERENCE_MODELS_GROUNDING_DINO_DEFAULT_MAX_DETECTIONS="200"

INFERENCE_MODELS_GROUNDING_DINO_DEFAULT_IOU_THRESHOLD Default: 0.5

export INFERENCE_MODELS_GROUNDING_DINO_DEFAULT_IOU_THRESHOLD="0.4"

Moondream2

INFERENCE_MODELS_MOONDREAM2_DEFAULT_MAX_NEW_TOKENS Default: 700

export INFERENCE_MODELS_MOONDREAM2_DEFAULT_MAX_NEW_TOKENS="1000"

OWLv2

INFERENCE_MODELS_OWLV2_DEFAULT_CONFIDENCE Default: 0.99

export INFERENCE_MODELS_OWLV2_DEFAULT_CONFIDENCE="0.95"

INFERENCE_MODELS_OWLV2_DEFAULT_IOU_THRESHOLD Default: Inherits from INFERENCE_MODELS_DEFAULT_IOU_THRESHOLD

export INFERENCE_MODELS_OWLV2_DEFAULT_IOU_THRESHOLD="0.4"

INFERENCE_MODELS_OWLV2_DEFAULT_MAX_DETECTIONS Default: Inherits from INFERENCE_MODELS_DEFAULT_MAX_DETECTIONS

export INFERENCE_MODELS_OWLV2_DEFAULT_MAX_DETECTIONS="200"

INFERENCE_MODELS_OWLV2_DEFAULT_CLASS_AGNOSTIC_NMS Default: Inherits from INFERENCE_MODELS_DEFAULT_CLASS_AGNOSTIC_NMS

export INFERENCE_MODELS_OWLV2_DEFAULT_CLASS_AGNOSTIC_NMS="true"

PaliGemma

INFERENCE_MODELS_PALIGEMMA_DEFAULT_MAX_NEW_TOKENS Default: 400

export INFERENCE_MODELS_PALIGEMMA_DEFAULT_MAX_NEW_TOKENS="500"

INFERENCE_MODELS_PALIGEMMA_DEFAULT_DO_SAMPLE Default: Inherits from INFERENCE_MODELS_DEFAULT_DO_SAMPLE

export INFERENCE_MODELS_PALIGEMMA_DEFAULT_DO_SAMPLE="true"

INFERENCE_MODELS_PALIGEMMA_DEFAULT_SKIP_SPECIAL_TOKENS Default: true

export INFERENCE_MODELS_PALIGEMMA_DEFAULT_SKIP_SPECIAL_TOKENS="false"

Gemma 4

These variables set the defaults used by the Gemma 4 Hugging Face implementation (Gemma4HF; see Gemma 4) for prompt(), pre_process_generation(), and generate() (including sampling kwargs when do_sample is true).

INFERENCE_MODELS_GEMMA4_DEFAULT_MAX_NEW_TOKENS Default: 512

export INFERENCE_MODELS_GEMMA4_DEFAULT_MAX_NEW_TOKENS="1024"

INFERENCE_MODELS_GEMMA4_DEFAULT_DO_SAMPLE Default: Inherits from INFERENCE_MODELS_DEFAULT_DO_SAMPLE

export INFERENCE_MODELS_GEMMA4_DEFAULT_DO_SAMPLE="true"

INFERENCE_MODELS_GEMMA4_DEFAULT_ENABLE_THINKING Default: false (passed to the processor chat template as enable_thinking)

export INFERENCE_MODELS_GEMMA4_DEFAULT_ENABLE_THINKING="true"

INFERENCE_MODELS_GEMMA4_DEFAULT_SKIP_SPECIAL_TOKENS Default: true

export INFERENCE_MODELS_GEMMA4_DEFAULT_SKIP_SPECIAL_TOKENS="false"

INFERENCE_MODELS_GEMMA4_DEFAULT_TEMPERATURE Default: 1.0 (used when do_sample is true)

export INFERENCE_MODELS_GEMMA4_DEFAULT_TEMPERATURE="0.8"

INFERENCE_MODELS_GEMMA4_DEFAULT_TOP_P Default: 0.95

export INFERENCE_MODELS_GEMMA4_DEFAULT_TOP_P="0.9"

INFERENCE_MODELS_GEMMA4_DEFAULT_TOP_K Default: 64

export INFERENCE_MODELS_GEMMA4_DEFAULT_TOP_K="32"

Qwen2.5-VL

INFERENCE_MODELS_QWEN25_VL_DEFAULT_MAX_NEW_TOKENS Default: 512

export INFERENCE_MODELS_QWEN25_VL_DEFAULT_MAX_NEW_TOKENS="1024"

INFERENCE_MODELS_QWEN25_VL_DEFAULT_DO_SAMPLE Default: Inherits from INFERENCE_MODELS_DEFAULT_DO_SAMPLE

export INFERENCE_MODELS_QWEN25_VL_DEFAULT_DO_SAMPLE="true"

INFERENCE_MODELS_QWEN25_VL_DEFAULT_SKIP_SPECIAL_TOKENS Default: true

export INFERENCE_MODELS_QWEN25_VL_DEFAULT_SKIP_SPECIAL_TOKENS="false"

Qwen3-VL

INFERENCE_MODELS_QWEN3_VL_DEFAULT_MAX_NEW_TOKENS Default: 512

export INFERENCE_MODELS_QWEN3_VL_DEFAULT_MAX_NEW_TOKENS="1024"

INFERENCE_MODELS_QWEN3_VL_DEFAULT_DO_SAMPLE Default: Inherits from INFERENCE_MODELS_DEFAULT_DO_SAMPLE

export INFERENCE_MODELS_QWEN3_VL_DEFAULT_DO_SAMPLE="true"

ResNet

INFERENCE_MODELS_RESNET_DEFAULT_CONFIDENCE Default: Inherits from INFERENCE_MODELS_DEFAULT_CONFIDENCE

export INFERENCE_MODELS_RESNET_DEFAULT_CONFIDENCE="0.5"

RF-DETR

INFERENCE_MODELS_RFDETR_DEFAULT_CONFIDENCE Default: Inherits from INFERENCE_MODELS_DEFAULT_CONFIDENCE

export INFERENCE_MODELS_RFDETR_DEFAULT_CONFIDENCE="0.5"

INFERENCE_MODELS_RFDETR_DEFAULT_MAX_DETECTIONS Default cap on RF-DETR instance-segmentation detections, applied by score BEFORE masks are upscaled to original resolution (bounds mask memory). Default: Inherits from INFERENCE_MODELS_DEFAULT_MAX_DETECTIONS

export INFERENCE_MODELS_RFDETR_DEFAULT_MAX_DETECTIONS="300"

The following variables select RF-DETR TensorRT pipeline implementations when a client cannot pass backend-specific from_pretrained arguments. Explicit arguments take precedence over these environment variables.

See Inference-Path Optimization Architecture for the selection model and the complete RF-DETR execution flow.

INFERENCE_MODELS_RFDETR_PREPROCESSOR Default: triton-universal-v1

Supported values: base, auto, threaded-exact-v1, triton-universal-v1.

export INFERENCE_MODELS_RFDETR_PREPROCESSOR="triton-universal-v1"

INFERENCE_MODELS_RFDETR_PREPROCESSOR_MAX_WORKERS Default: 4

Controls the bounded worker count used by threaded-exact-v1.

export INFERENCE_MODELS_RFDETR_PREPROCESSOR_MAX_WORKERS="4"

INFERENCE_MODELS_RFDETR_POSTPROCESSOR Default: triton-fused-v1

Supported values: base, auto, triton-fused-v1.

export INFERENCE_MODELS_RFDETR_POSTPROCESSOR="triton-fused-v1"

Code that can pass backend-specific arguments may instead provide a composed, immutable execution plan:

from inference_models import AutoModel
from inference_models.models.rfdetr.optimization.execution_plan import (
    RFDetrExecutionPlan,
)

plan = RFDetrExecutionPlan(
    preprocessor_id="triton-universal-v1",
    postprocessor_id="triton-fused-v1",
    allow_compatibility_fallback=True,
)
model = AutoModel.from_pretrained(
    "rfdetr-small",
    backend="trt",
    rfdetr_execution_plan=plan,
)

Public preprocessing synchronizes by default, so its result can be consumed by an independent forward() call without relying on model-owned readiness state:

model = AutoModel.from_pretrained(
    "rfdetr-small",
    backend="trt",
    rfdetr_execution_plan=plan,
)
preprocessed, metadata = model.pre_process(image)
raw_predictions = model.forward(preprocessed)

This invocation-boundary policy is intentionally separate from the execution plan. Composed model(...) and infer() calls pass independent_stage_execution=False to preprocessing internally, record a CUDA event, and let forward() wait on that event without a host synchronization.

The plan also reserves independently selectable buffer-strategy, scheduler, and engine plugin stages. Those stages currently accept only base. When supplied, an explicit plan takes precedence and the implementation-selection environment variables are not read.

When a selected optimized stage declares that it cannot preserve a model or request contract, RF-DETR uses its declared base fallback and records the requested implementation, effective implementation, and reason in logs and runtime metadata. This policy applies consistently to preprocessing and postprocessing. Set allow_compatibility_fallback=False in an explicit plan to require the selected implementation or an error. Compilation, CUDA, allocation, and other execution failures are never converted into fallbacks.

An all-False PreProcessingOverrides object is a no-op and remains compatible with triton-universal-v1. Requests with any active preprocessing override use the declared base fallback. A distinct request-level fallback reason is warned once per model instance rather than once per inference.

When Triton is unavailable, uint8 universal preprocessing and fused postprocessing declare a compatibility miss and use their base fallback. Floating-point tensor preprocessing remains eligible for triton-universal-v1 because that input path uses Torch operations and does not require Triton kernels.

Roboflow Instant

INFERENCE_MODELS_ROBOFLOW_INSTANT_DEFAULT_CONFIDENCE Default: 0.99

export INFERENCE_MODELS_ROBOFLOW_INSTANT_DEFAULT_CONFIDENCE="0.95"

INFERENCE_MODELS_ROBOFLOW_INSTANT_DEFAULT_IOU_THRESHOLD Default: 0.3

export INFERENCE_MODELS_ROBOFLOW_INSTANT_DEFAULT_IOU_THRESHOLD="0.4"

INFERENCE_MODELS_ROBOFLOW_INSTANT_MAX_DETECTIONS Default: Inherits from INFERENCE_MODELS_DEFAULT_MAX_DETECTIONS

export INFERENCE_MODELS_ROBOFLOW_INSTANT_MAX_DETECTIONS="200"

SmolVLM

INFERENCE_MODELS_SMOL_VLM_DEFAULT_MAX_NEW_TOKENS Default: 400

export INFERENCE_MODELS_SMOL_VLM_DEFAULT_MAX_NEW_TOKENS="500"

INFERENCE_MODELS_SMOL_VLM_DEFAULT_DO_SAMPLE Default: Inherits from INFERENCE_MODELS_DEFAULT_DO_SAMPLE

export INFERENCE_MODELS_SMOL_VLM_DEFAULT_DO_SAMPLE="true"

INFERENCE_MODELS_SMOL_VLM_DEFAULT_SKIP_SPECIAL_TOKENS Default: true

export INFERENCE_MODELS_SMOL_VLM_DEFAULT_SKIP_SPECIAL_TOKENS="false"

ViT

INFERENCE_MODELS_VIT_CLASSIFIER_DEFAULT_CONFIDENCE Default: Inherits from INFERENCE_MODELS_DEFAULT_CONFIDENCE

export INFERENCE_MODELS_VIT_CLASSIFIER_DEFAULT_CONFIDENCE="0.5"

YOLACT

INFERENCE_MODELS_YOLACT_DEFAULT_CONFIDENCE Default: Inherits from INFERENCE_MODELS_DEFAULT_CONFIDENCE

export INFERENCE_MODELS_YOLACT_DEFAULT_CONFIDENCE="0.5"

INFERENCE_MODELS_YOLACT_DEFAULT_IOU_THRESHOLD Default: Inherits from INFERENCE_MODELS_DEFAULT_IOU_THRESHOLD

export INFERENCE_MODELS_YOLACT_DEFAULT_IOU_THRESHOLD="0.4"

INFERENCE_MODELS_YOLACT_DEFAULT_MAX_DETECTIONS Default: Inherits from INFERENCE_MODELS_DEFAULT_MAX_DETECTIONS

export INFERENCE_MODELS_YOLACT_DEFAULT_MAX_DETECTIONS="200"

INFERENCE_MODELS_YOLACT_DEFAULT_CLASS_AGNOSTIC_NMS Default: Inherits from INFERENCE_MODELS_DEFAULT_CLASS_AGNOSTIC_NMS

export INFERENCE_MODELS_YOLACT_DEFAULT_CLASS_AGNOSTIC_NMS="true"

YOLO-NAS

INFERENCE_MODELS_YOLONAS_DEFAULT_CONFIDENCE Default: Inherits from INFERENCE_MODELS_DEFAULT_CONFIDENCE

export INFERENCE_MODELS_YOLONAS_DEFAULT_CONFIDENCE="0.5"

INFERENCE_MODELS_YOLONAS_DEFAULT_IOU_THRESHOLD Default: Inherits from INFERENCE_MODELS_DEFAULT_IOU_THRESHOLD

export INFERENCE_MODELS_YOLONAS_DEFAULT_IOU_THRESHOLD="0.4"

INFERENCE_MODELS_YOLONAS_DEFAULT_MAX_DETECTIONS Default: Inherits from INFERENCE_MODELS_DEFAULT_MAX_DETECTIONS

export INFERENCE_MODELS_YOLONAS_DEFAULT_MAX_DETECTIONS="200"

INFERENCE_MODELS_YOLONAS_DEFAULT_CLASS_AGNOSTIC_NMS Default: Inherits from INFERENCE_MODELS_DEFAULT_CLASS_AGNOSTIC_NMS

export INFERENCE_MODELS_YOLONAS_DEFAULT_CLASS_AGNOSTIC_NMS="true"

YOLOv5

INFERENCE_MODELS_YOLOV5_DEFAULT_CONFIDENCE Default: Inherits from INFERENCE_MODELS_DEFAULT_CONFIDENCE

export INFERENCE_MODELS_YOLOV5_DEFAULT_CONFIDENCE="0.5"

INFERENCE_MODELS_YOLOV5_DEFAULT_IOU_THRESHOLD Default: Inherits from INFERENCE_MODELS_DEFAULT_IOU_THRESHOLD

export INFERENCE_MODELS_YOLOV5_DEFAULT_IOU_THRESHOLD="0.4"

INFERENCE_MODELS_YOLOV5_DEFAULT_MAX_DETECTIONS Default: Inherits from INFERENCE_MODELS_DEFAULT_MAX_DETECTIONS

export INFERENCE_MODELS_YOLOV5_DEFAULT_MAX_DETECTIONS="200"

INFERENCE_MODELS_YOLOV5_DEFAULT_CLASS_AGNOSTIC_NMS Default: Inherits from INFERENCE_MODELS_DEFAULT_CLASS_AGNOSTIC_NMS

export INFERENCE_MODELS_YOLOV5_DEFAULT_CLASS_AGNOSTIC_NMS="true"

YOLOv7

INFERENCE_MODELS_YOLOV7_DEFAULT_CONFIDENCE Default: Inherits from INFERENCE_MODELS_DEFAULT_CONFIDENCE

export INFERENCE_MODELS_YOLOV7_DEFAULT_CONFIDENCE="0.5"

INFERENCE_MODELS_YOLOV7_DEFAULT_IOU_THRESHOLD Default: Inherits from INFERENCE_MODELS_DEFAULT_IOU_THRESHOLD

export INFERENCE_MODELS_YOLOV7_DEFAULT_IOU_THRESHOLD="0.4"

INFERENCE_MODELS_YOLOV7_DEFAULT_MAX_DETECTIONS Default: Inherits from INFERENCE_MODELS_DEFAULT_MAX_DETECTIONS

export INFERENCE_MODELS_YOLOV7_DEFAULT_MAX_DETECTIONS="200"

INFERENCE_MODELS_YOLOV7_DEFAULT_CLASS_AGNOSTIC_NMS Default: Inherits from INFERENCE_MODELS_DEFAULT_CLASS_AGNOSTIC_NMS

export INFERENCE_MODELS_YOLOV7_DEFAULT_CLASS_AGNOSTIC_NMS="true"

YOLOv8/v9/v11/v12 (Ultralytics)

INFERENCE_MODELS_YOLO_ULTRALYTICS_DEFAULT_CONFIDENCE Default: Inherits from INFERENCE_MODELS_DEFAULT_CONFIDENCE

export INFERENCE_MODELS_YOLO_ULTRALYTICS_DEFAULT_CONFIDENCE="0.5"

INFERENCE_MODELS_YOLO_ULTRALYTICS_DEFAULT_IOU_THRESHOLD Default: Inherits from INFERENCE_MODELS_DEFAULT_IOU_THRESHOLD

export INFERENCE_MODELS_YOLO_ULTRALYTICS_DEFAULT_IOU_THRESHOLD="0.4"

INFERENCE_MODELS_YOLO_ULTRALYTICS_DEFAULT_MAX_DETECTIONS Default: Inherits from INFERENCE_MODELS_DEFAULT_MAX_DETECTIONS

export INFERENCE_MODELS_YOLO_ULTRALYTICS_DEFAULT_MAX_DETECTIONS="200"

INFERENCE_MODELS_YOLO_ULTRALYTICS_DEFAULT_CLASS_AGNOSTIC_NMS Default: Inherits from INFERENCE_MODELS_DEFAULT_CLASS_AGNOSTIC_NMS

export INFERENCE_MODELS_YOLO_ULTRALYTICS_DEFAULT_CLASS_AGNOSTIC_NMS="true"

INFERENCE_MODELS_YOLO_ULTRALYTICS_DEFAULT_KEY_POINTS_THRESHOLD Default: 0.0

export INFERENCE_MODELS_YOLO_ULTRALYTICS_DEFAULT_KEY_POINTS_THRESHOLD="0.4"

YOLOv10

INFERENCE_MODELS_YOLOV10_DEFAULT_CONFIDENCE Default: Inherits from INFERENCE_MODELS_DEFAULT_CONFIDENCE

export INFERENCE_MODELS_YOLOV10_DEFAULT_CONFIDENCE="0.5"

INFERENCE_MODELS_YOLOV10_DEFAULT_MAX_DETECTIONS Default: Inherits from INFERENCE_MODELS_DEFAULT_MAX_DETECTIONS

export INFERENCE_MODELS_YOLOV10_DEFAULT_MAX_DETECTIONS="200"

Logging

LOG_LEVEL
Set the log level for the library. Default: WARNING

export LOG_LEVEL="DEBUG"

VERBOSE_LOG_LEVEL
Set the log level for verbose logging. Default: INFO

export VERBOSE_LOG_LEVEL="DEBUG"

DISABLE_VERBOSE_LOGGER
Disable verbose logging. Default: false

export DISABLE_VERBOSE_LOGGER="true"

DISABLE_INTERACTIVE_PROGRESS_BARS
Disable interactive progress bars. Default: false

export DISABLE_INTERACTIVE_PROGRESS_BARS="true"

Advanced Configuration

SAM3

INFERENCE_MODELS_SAM3_MASK_PROCESSING_CHUNK_SIZE Number of SAM3 concept-segmentation masks upscaled to original resolution and encoded per slice. Bounds postprocessing memory (working set is chunk x H x W float32 instead of detections x H x W). Must be >= 1. Default: 8

export INFERENCE_MODELS_SAM3_MASK_PROCESSING_CHUNK_SIZE="8"

Input Validation

ALLOW_URL_INPUT Allow URLs as image input. Used by models like OWL-V2, when access to larger datasets provided as references is needed. Default: true

export ALLOW_URL_INPUT="true"

ALLOW_NON_HTTPS_URL_INPUT Allow non-HTTPS URLs. Used by models like OWL-V2, when access to larger datasets provided as references is needed. Default: false

export ALLOW_NON_HTTPS_URL_INPUT="true"  # Use with caution

ALLOW_URL_INPUT_WITHOUT_FQDN Allow URLs without FQDN. Used by models like OWL-V2, when access to larger datasets provided as references is needed. Default: false

export ALLOW_URL_INPUT_WITHOUT_FQDN="true"  # Use with caution

WHITELISTED_DESTINATIONS_FOR_URL_INPUT
Comma-separated list of allowed destinations for URL input. Used by models like OWL-V2, when access to larger datasets provided as references is needed. Default: None

export WHITELISTED_DESTINATIONS_FOR_URL_INPUT="google.com,github.com"

BLACKLISTED_DESTINATIONS_FOR_URL_INPUT
Comma-separated list of allowed destinations for URL input. Used by models like OWL-V2, when access to larger datasets provided as references is needed. Default: None

export BLACKLISTED_DESTINATIONS_FOR_URL_INPUT="google.com,github.com"

ALLOW_LOCAL_STORAGE_ACCESS_FOR_REFERENCE_DATA
Allow local storage access for reference data. Used by models like OWL-V2, when access to larger datasets provided as references is needed. Default: true

export ALLOW_LOCAL_STORAGE_ACCESS_FOR_REFERENCE_DATA="true"