Primary use stated or inferred cautiously from official documentation.
N
NVIDIA / nvidia/cdf-sim2real-smt-solderlight
CDF Sim2Real SMT Solderlight
NVIDIA image-editing model, version 1.0.0 released on October 20, 2026. It transforms synthetic PCB component crops rendered in Omniverse into an optical-inspection lighting style. It derives from Qwen-Image-Edit with a fused LoRA fine-tune: the package includes the complete pipeline, not a separate adapter. Users supply the image; the English instruction is fixed, not a free-form prompt.
01 / SUMMARY
The essential
Maximum input capacity when published by the source.
Documented maximum generation limit.
Verified availability channels.
Primary functional family and verified specialties.
Declared terms for API access, model weights, or self-hosting.
02 / TECHNICAL SHEET
Limits and integration
| API ID | nvidia/cdf-sim2real-smt-solderlight |
|---|---|
| Model type | Image |
| Access model | Open weights |
| License | OpenMDW 1.1 |
| Deployment | Self-hosted deployment with NVIDIA GPUs |
| Release | 20 OCT 2026 |
| Knowledge cutoff | Not published |
| Entrance | Image · Text |
| Exit | Image |
| Context window | One RGB PCB component image and a fixed English instruction |
| maximum output | RGB PNG image; target area of 262,144 pixels |
| Reasoning | Lighting edits, not automated inspection |
| Published tools | |
| Structured Outings | Not published |
| Batch processing | Not published |
| Prompt cache | Not published |
| Fine-tuning | Not published |
| Verified platforms | Hugging Face · PyTorch · Diffusers |
Sim-to-real data augmentation for developing PCB inspection systems. It aims to preserve component geometry, markings and shape while changing lighting. Load it from the model subfolder with QwenImageEditPipeline. It targets crops similar to its training distribution, not general photography, text-to-image generation or defect diagnosis.
Training uses an area of 262,144 pixels, equivalent to 512 × 512 for square crops: about 0.262 megapixels, not 0.5. Other resolutions may degrade style fidelity. It is not a PCB quality decision-maker: acceptance or rejection requires a downstream system and human review. Inferama has neither run the model nor verified its industrial performance.
03 / CAPABILITIES
What it can do
Orientation
Sim-to-real data augmentation for developing PCB inspection systems. It aims to preserve component geometry, markings and shape while changing lighting. Load it from the model subfolder with QwenImageEditPipeline. It targets crops similar to its training distribution, not general photography, text-to-image generation or defect diagnosis.
Context
One RGB PCB component image and a fixed English instruction · RGB PNG image; target area of 262,144 pixels
Tools and integration
Not published
Access
Hugging Face · downloadable checkpoint
Input modalities
ImageText
04 / PRICES
Documented cost
| Concept | Worth | Unit/condition |
|---|---|---|
| Standard input | No published official API rate | Standard API rate |
| Cached input | Not published | Reading reused prefixes |
| Cache write or storage | Not published | The condition varies by provider |
| Standard output | Self-hosted infrastructure cost | May include reasoning tokens |
| Batch input | Not published | Asynchronous processing |
| Batch output | Not published | Asynchronous processing |
Public weights under OpenMDW 1.1; original Qwen-Image-Edit components retain Apache 2.0. The model card estimates about 20B parameters and 170M trained, fused LoRA parameters. We have verified neither a commercial API nor image or token pricing: numeric costs remain empty. Budget for GPU, memory, storage and self-hosted operations.
i Prices change and may depend on level, region or context length. Check the source before making a decision.
05 / EVALUATIONS
How to read the results
“
NVIDIA describes 364 training pairs and qualitative visual evaluation, with no independent validation or test split or packaged real-photograph reference set. It publishes no numerical industrial-quality scores. We do not inherit base-model benchmarks. For real-world evaluation, separate training and test boards, measure geometry preservation and introduced defects, and validate the impact on the downstream inspector with human review.
i Qualitative vendor evaluation; no verified numerical scores.
Inferama only highlights a “best result” when the metric, test set, configuration, and date allow for an equivalent comparison. The supplier's figures are presented as claims from its own source.
06 / VERSIONS
Chronology
v1.0.0
Release date stated in the official model card; do not confuse it with repository creation or modification.
07 / SOURCES