Instructions to use Subh775/Dis-Seg-Former with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use Subh775/Dis-Seg-Former with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("image-segmentation", model="Subh775/Dis-Seg-Former")# Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("Subh775/Dis-Seg-Former", dtype="auto") - Notebooks
- Google Colab
- Kaggle
| { | |
| "train_config": { | |
| "lr": 0.0001, | |
| "lr_encoder": 0.00015, | |
| "batch_size": 4, | |
| "grad_accum_steps": 2, | |
| "auto_batch_target_effective": 16, | |
| "auto_batch_max_targets_per_image": 100, | |
| "auto_batch_ema_headroom": 0.7, | |
| "epochs": 20, | |
| "resume": null, | |
| "ema_decay": 0.993, | |
| "ema_tau": 100, | |
| "lr_drop": 100, | |
| "checkpoint_interval": 10, | |
| "skip_best_epochs": 0, | |
| "warmup_epochs": 0.0, | |
| "lr_vit_layer_decay": 0.8, | |
| "lr_component_decay": 0.7, | |
| "drop_path": 0.0, | |
| "group_detr": 13, | |
| "ia_bce_loss": true, | |
| "cls_loss_coef": 5.0, | |
| "num_select": null, | |
| "dataset_file": "roboflow", | |
| "square_resize_div_64": true, | |
| "dataset_dir": "/kaggle/working/leaf-disease-seg-1", | |
| "output_dir": "output", | |
| "multi_scale": true, | |
| "expanded_scales": true, | |
| "do_random_resize_via_padding": false, | |
| "use_ema": true, | |
| "ema_update_interval": 1, | |
| "num_workers": 2, | |
| "weight_decay": 0.0001, | |
| "early_stopping": true, | |
| "early_stopping_patience": 3, | |
| "early_stopping_min_delta": 0.001, | |
| "early_stopping_use_ema": false, | |
| "progress_bar": "rich", | |
| "tensorboard": true, | |
| "wandb": true, | |
| "mlflow": false, | |
| "clearml": false, | |
| "project": "Agrivision", | |
| "run": "rfdetr-seg-nano-ep20", | |
| "class_names": null, | |
| "run_test": false, | |
| "segmentation_head": true, | |
| "eval_max_dets": 500, | |
| "eval_interval": 1, | |
| "log_per_class_metrics": true, | |
| "aug_config": null, | |
| "augmentation_backend": "cpu", | |
| "save_dataset_grids": false, | |
| "notes": null, | |
| "accelerator": "auto", | |
| "clip_max_norm": 0.1, | |
| "seed": null, | |
| "sync_bn": false, | |
| "strategy": "ddp_find_unused_parameters_true", | |
| "devices": "auto", | |
| "num_nodes": 1, | |
| "fp16_eval": false, | |
| "lr_scheduler": "step", | |
| "lr_min_factor": 0.0, | |
| "dont_save_weights": false, | |
| "train_log_sync_dist": false, | |
| "train_log_on_step": false, | |
| "compute_val_loss": true, | |
| "compute_test_loss": true, | |
| "pin_memory": null, | |
| "persistent_workers": null, | |
| "prefetch_factor": null, | |
| "mask_point_sample_ratio": 16, | |
| "mask_ce_loss_coef": 5.0, | |
| "mask_dice_loss_coef": 5.0 | |
| }, | |
| "model_config": { | |
| "encoder": "dinov2_windowed_small", | |
| "out_feature_indexes": [ | |
| 3, | |
| 6, | |
| 9, | |
| 12 | |
| ], | |
| "dec_layers": 4, | |
| "two_stage": true, | |
| "projector_scale": [ | |
| "P4" | |
| ], | |
| "hidden_dim": 256, | |
| "patch_size": 12, | |
| "num_windows": 1, | |
| "sa_nheads": 8, | |
| "ca_nheads": 16, | |
| "dec_n_points": 2, | |
| "num_queries": 100, | |
| "num_select": 100, | |
| "bbox_reparam": true, | |
| "lite_refpoint_refine": true, | |
| "layer_norm": true, | |
| "amp": true, | |
| "num_channels": 3, | |
| "num_classes": 2, | |
| "pretrain_weights": "/root/.roboflow/models/rf-detr-seg-nano.pt", | |
| "device": "cuda", | |
| "resolution": 312, | |
| "group_detr": 13, | |
| "gradient_checkpointing": false, | |
| "compile": false, | |
| "fused_optimizer": true, | |
| "positional_encoding_size": 26, | |
| "ia_bce_loss": true, | |
| "cls_loss_coef": 1.0, | |
| "segmentation_head": true, | |
| "mask_downsample_ratio": 4, | |
| "backbone_lora": false, | |
| "freeze_encoder": false, | |
| "license": "Apache-2.0", | |
| "model_name": "RFDETRSegNano" | |
| }, | |
| "model_config_type": "RFDETRSegNanoConfig", | |
| "class_names": [ | |
| "Leaf", | |
| "Leaf Diseases" | |
| ], | |
| "num_classes": 2 | |
| } |