Added convert_from_recipe script #1568

BloodAxe · 2023-10-24T11:10:47Z

This PR adds convert_from_recipe entry point to create self-contained script to train a model with given recipe.
Generate script file contains full config paramteres to train the model.

Example usage:

General use: python -m super_gradients.convert_recipe_to_code --config-name="DESIRED_RECIPE".

How it works

We compose the recipe config using hydra.compose but we do no instantiate objects using hydra.
We traverse config recursively to identify places, where hydra instantiation takes place
We replace hydra instantiation with variable references & instantiate those variables in generated code. This is the most tricky part of the script.
We run black formatter on generated content file to make it look pretty and dump to output file

Open questions

Generated config is pretty verbose (expected). Should we somehow attempt to "shrink" it and drop properties whose values are the default ones?
I'm not particularly proud how _lr_updates = super_gradients.training.utils.utils.empty_list() looks like. Fixing such special cases would require adding if checks for each such special case.
How should we test this. Would a test that runs over all existing recipes and exporting them work?

Example of the generated script file:

import super_gradients
from super_gradients import init_trainer, Trainer
from super_gradients.training.utils.distributed_training_utils import setup_device
from super_gradients.training import models, dataloaders
from super_gradients.common.data_types.enum import MultiGPUMode, StrictLoad
import numpy as np


def main():
    init_trainer()
    setup_device(device=None, multi_gpu="DDP", num_gpus=8)

    trainer = Trainer(experiment_name="coco2017_yolo_nas_s", ckpt_root_dir="None")

    num_classes = 80
    arch_params = {
        "in_channels": 3,
        "backbone": {
            "NStageBackbone": {
                "stem": {"YoloNASStem": {"out_channels": 48}},
                "stages": [
                    {"YoloNASStage": {"out_channels": 96, "num_blocks": 2, "activation_type": "relu", "hidden_channels": 32, "concat_intermediates": False}},
                    {"YoloNASStage": {"out_channels": 192, "num_blocks": 3, "activation_type": "relu", "hidden_channels": 64, "concat_intermediates": False}},
                    {"YoloNASStage": {"out_channels": 384, "num_blocks": 5, "activation_type": "relu", "hidden_channels": 96, "concat_intermediates": False}},
                    {"YoloNASStage": {"out_channels": 768, "num_blocks": 2, "activation_type": "relu", "hidden_channels": 192, "concat_intermediates": False}},
                ],
                "context_module": {"SPP": {"output_channels": 768, "activation_type": "relu", "k": [5, 9, 13]}},
                "out_layers": ["stage1", "stage2", "stage3", "context_module"],
            }
        },
        "neck": {
            "YoloNASPANNeckWithC2": {
                "neck1": {
                    "YoloNASUpStage": {
                        "out_channels": 192,
                        "num_blocks": 2,
                        "hidden_channels": 64,
                        "width_mult": 1,
                        "depth_mult": 1,
                        "activation_type": "relu",
                        "reduce_channels": True,
                    }
                },
                "neck2": {
                    "YoloNASUpStage": {
                        "out_channels": 96,
                        "num_blocks": 2,
                        "hidden_channels": 48,
                        "width_mult": 1,
                        "depth_mult": 1,
                        "activation_type": "relu",
                        "reduce_channels": True,
                    }
                },
                "neck3": {
                    "YoloNASDownStage": {
                        "out_channels": 192,
                        "num_blocks": 2,
                        "hidden_channels": 64,
                        "activation_type": "relu",
                        "width_mult": 1,
                        "depth_mult": 1,
                    }
                },
                "neck4": {
                    "YoloNASDownStage": {
                        "out_channels": 384,
                        "num_blocks": 2,
                        "hidden_channels": 64,
                        "activation_type": "relu",
                        "width_mult": 1,
                        "depth_mult": 1,
                    }
                },
            }
        },
        "heads": {
            "NDFLHeads": {
                "num_classes": 80,
                "reg_max": 16,
                "heads_list": [
                    {"YoloNASDFLHead": {"inter_channels": 128, "width_mult": 0.5, "first_conv_group_size": 0, "stride": 8}},
                    {"YoloNASDFLHead": {"inter_channels": 256, "width_mult": 0.5, "first_conv_group_size": 0, "stride": 16}},
                    {"YoloNASDFLHead": {"inter_channels": 512, "width_mult": 0.5, "first_conv_group_size": 0, "stride": 32}},
                ],
            }
        },
        "bn_eps": 0.001,
        "bn_momentum": 0.03,
        "inplace_act": True,
        "_convert_": "all",
        "num_classes": 80,
    }

    model = models.get(
        model_name="yolo_nas_s",
        num_classes=num_classes,
        arch_params=arch_params,
        strict_load=StrictLoad.NO_KEY_MATCHING,
        pretrained_weights=None,
        checkpoint_path=None,
        load_backbone=False,
        checkpoint_num_classes=None,
    )

    train_dataloader = dataloaders.get(
        name="coco2017_train_yolo_nas",
        dataset_params={
            "data_dir": "/data/coco",
            "subdir": "images/train2017",
            "json_file": "instances_train2017.json",
            "input_dim": [640, 640],
            "cache_dir": None,
            "cache": False,
            "cache_annotations": True,
            "ignore_empty_annotations": True,
            "transforms": [
                {
                    "DetectionRandomAffine": {
                        "degrees": 0,
                        "translate": 0.25,
                        "scales": [0.5, 1.5],
                        "shear": 0.0,
                        "target_size": None,
                        "filter_box_candidates": True,
                        "wh_thr": 2,
                        "area_thr": 0.1,
                        "ar_thr": 20,
                    }
                },
                {"DetectionRGB2BGR": {"prob": 0.5}},
                {"DetectionHSV": {"prob": 0.5, "hgain": 18, "sgain": 30, "vgain": 30}},
                {"DetectionHorizontalFlip": {"prob": 0.5}},
                {"DetectionMixup": {"input_dim": None, "mixup_scale": [0.5, 1.5], "prob": 0.5, "flip_prob": 0.5}},
                {"DetectionPaddedRescale": {"input_dim": [640, 640], "pad_value": 114}},
                {"DetectionStandardize": {"max_value": 255.0}},
                {"DetectionTargetsFormatTransform": {"output_format": "LABEL_CXCYWH"}},
            ],
            "tight_box_rotation": False,
            "class_inclusion_list": None,
            "max_num_samples": None,
            "with_crowd": False,
        },
        dataloader_params={"batch_size": 32, "num_workers": 8, "shuffle": True, "drop_last": True, "pin_memory": True, "collate_fn": "DetectionCollateFN"},
    )

    val_dataloader = dataloaders.get(
        name="coco2017_val_yolo_nas",
        dataset_params={
            "data_dir": "/data/coco",
            "subdir": "images/val2017",
            "json_file": "instances_val2017.json",
            "input_dim": [636, 636],
            "cache_dir": None,
            "cache": False,
            "cache_annotations": True,
            "ignore_empty_annotations": True,
            "transforms": [
                {"DetectionRGB2BGR": {"prob": 1}},
                {"DetectionPadToSize": {"output_size": [640, 640], "pad_value": 114}},
                {"DetectionStandardize": {"max_value": 255.0}},
                "DetectionImagePermute",
                {"DetectionTargetsFormatTransform": {"input_dim": [640, 640], "output_format": "LABEL_CXCYWH"}},
            ],
            "tight_box_rotation": False,
            "class_inclusion_list": None,
            "max_num_samples": None,
            "with_crowd": True,
        },
        dataloader_params={
            "batch_size": 25,
            "num_workers": 8,
            "drop_last": False,
            "shuffle": False,
            "pin_memory": True,
            "collate_fn": "CrowdDetectionCollateFN",
        },
    )

    _lr_updates = super_gradients.training.utils.utils.empty_list()

    _valid_metrics_list_0_detectionmetrics_post_prediction_callback = super_gradients.training.models.detection_models.pp_yolo_e.PPYoloEPostPredictionCallback(
        score_threshold=0.01, nms_top_k=1000, max_predictions=300, nms_threshold=0.7
    )

    training_hyperparams = {
        "resume": None,
        "run_id": None,
        "resume_path": None,
        "resume_from_remote_sg_logger": False,
        "ckpt_name": "ckpt_latest.pth",
        "lr_mode": "CosineLRScheduler",
        "lr_schedule_function": None,
        "lr_warmup_epochs": 0,
        "lr_warmup_steps": 1000,
        "lr_cooldown_epochs": 0,
        "warmup_initial_lr": 1e-06,
        "step_lr_update_freq": None,
        "cosine_final_lr_ratio": 0.1,
        "warmup_mode": "LinearBatchLRWarmup",
        "lr_updates": _lr_updates,
        "pre_prediction_callback": None,
        "optimizer": "AdamW",
        "optimizer_params": {"weight_decay": 1e-05},
        "load_opt_params": True,
        "zero_weight_decay_on_bias_and_bn": True,
        "loss": {"PPYoloELoss": {"use_static_assigner": False, "num_classes": 80, "reg_max": 16}},
        "criterion_params": {},
        "ema": True,
        "ema_params": {"decay": 0.9997, "decay_type": "threshold", "beta": 15},
        "train_metrics_list": [],
        "valid_metrics_list": [
            {
                "DetectionMetrics": {
                    "score_thres": 0.1,
                    "top_k_predictions": 300,
                    "num_cls": 80,
                    "normalize_targets": True,
                    "post_prediction_callback": _valid_metrics_list_0_detectionmetrics_post_prediction_callback,
                }
            }
        ],
        "metric_to_watch": "mAP@0.50:0.95",
        "greater_metric_to_watch_is_better": True,
        "launch_tensorboard": False,
        "tensorboard_port": None,
        "tb_files_user_prompt": False,
        "save_tensorboard_to_s3": False,
        "precise_bn": False,
        "precise_bn_batch_size": None,
        "sync_bn": True,
        "silent_mode": False,
        "mixed_precision": True,
        "save_ckpt_epoch_list": [100, 200, 250],
        "average_best_models": True,
        "dataset_statistics": False,
        "batch_accumulate": 1,
        "run_validation_freq": 1,
        "run_test_freq": 1,
        "save_model": True,
        "seed": 42,
        "phase_callbacks": [],
        "log_installed_packages": True,
        "clip_grad_norm": None,
        "ckpt_best_name": "ckpt_best.pth",
        "max_train_batches": None,
        "max_valid_batches": None,
        "sg_logger": "base_sg_logger",
        "sg_logger_params": {
            "tb_files_user_prompt": False,
            "launch_tensorboard": False,
            "tensorboard_port": None,
            "save_checkpoints_remote": False,
            "save_tensorboard_remote": False,
            "save_logs_remote": False,
            "monitor_system": True,
        },
        "torch_compile": False,
        "torch_compile_loss": False,
        "torch_compile_options": {"mode": "reduce-overhead", "fullgraph": False, "dynamic": False, "backend": "inductor", "options": None, "disable": False},
        "_convert_": "all",
        "max_epochs": 300,
        "initial_lr": 0.0002,
    }

    # TRAIN
    result = trainer.train(
        model=model,
        train_loader=train_dataloader,
        valid_loader=val_dataloader,
        training_params=training_hyperparams,
    )

    print(result)


if __name__ == "__main__":
    main()

src/super_gradients/convert_from_recipe.py

# Conflicts: # tests/deci_core_unit_test_suite_runner.py

tests/unit_tests/convert_from_recipe_tests.py

shaydeci · 2023-10-26T13:18:56Z

This PR adds convert_from_recipe entry point to create self-contained script to train a model with given recipe. Generate script file contains full config paramteres to train the model.

Example usage:

General use: python -m super_gradients.convert_from_recipe --config-name="DESIRED_RECIPE".

How it works

We compose the recipe config using hydra.compose but we do no instantiate objects using hydra.
We traverse config recursively to identify places, where hydra instantiation takes place
We replace hydra instantiation with variable references & instantiate those variables in generated code. This is the most tricky part of the script.
We run black formatter on generated content file to make it look pretty and dump to output file

Open questions

Generated config is pretty verbose (expected). Should we somehow attempt to "shrink" it and drop properties whose values are the default ones?
I'm not particularly proud how _lr_updates = super_gradients.training.utils.utils.empty_list() looks like. Fixing such special cases would require adding if checks for each such special case.
How should we test this. Would a test that runs over all existing recipes and exporting them work?

Example of the generated script file:

import super_gradients
from super_gradients import init_trainer, Trainer
from super_gradients.training.utils.distributed_training_utils import setup_device
from super_gradients.training import models, dataloaders
from super_gradients.common.data_types.enum import MultiGPUMode, StrictLoad
import numpy as np


def main():
    init_trainer()
    setup_device(device=None, multi_gpu="DDP", num_gpus=8)

    trainer = Trainer(experiment_name="coco2017_pose_dekr_w32_no_dc", ckpt_root_dir="None")

    num_classes = 17
    arch_params = {
        "SPEC": {
            "FINAL_CONV_KERNEL": 1,
            "STAGES": {
                "NUM_STAGES": 3,
                "NUM_MODULES": [1, 4, 3],
                "NUM_BRANCHES": [2, 3, 4],
                "BLOCK": ["BASIC", "BASIC", "BASIC"],
                "NUM_BLOCKS": [[4, 4], [4, 4, 4], [4, 4, 4, 4]],
                "NUM_CHANNELS": [[32, 64], [32, 64, 128], [32, 64, 128, 256]],
                "FUSE_METHOD": ["SUM", "SUM", "SUM"],
            },
            "HEAD_HEATMAP": {
                "BLOCK": "BASIC",
                "NUM_BLOCKS": 1,
                "NUM_CHANNELS": 32,
                "DILATION_RATE": 1,
                "HEATMAP_APPLY_SIGMOID": False,
            },
            "HEAD_OFFSET": {"BLOCK": "BASIC", "DILATION_RATE": 5, "NUM_BLOCKS": 2, "NUM_CHANNELS_PERKPT": 15},
        },
        "num_classes": 17,
    }

    model = models.get(
        model_name="dekr_w32_no_dc",
        num_classes=num_classes,
        arch_params=arch_params,
        strict_load=StrictLoad.KEY_MATCHING,
        pretrained_weights=None,
        checkpoint_path=None,
        load_backbone=False,
        checkpoint_num_classes=None,
    )

    train_dataloader = dataloaders.get(
        name="coco2017_pose_train",
        dataset_params={
            "data_dir": "/data/coco",
            "images_dir": "images/train2017",
            "json_file": "annotations/person_keypoints_train2017.json",
            "include_empty_samples": False,
            "min_instance_area": 64,
            "edge_links": [
                [0, 1],
                [0, 2],
                [1, 2],
                [1, 3],
                [2, 4],
                [3, 5],
                [4, 6],
                [5, 6],
                [5, 7],
                [5, 11],
                [6, 8],
                [6, 12],
                [7, 9],
                [8, 10],
                [11, 12],
                [11, 13],
                [12, 14],
                [13, 15],
                [14, 16],
            ],
            "edge_colors": [
                [214, 39, 40],
                [148, 103, 189],
                [44, 160, 44],
                [140, 86, 75],
                [227, 119, 194],
                [127, 127, 127],
                [188, 189, 34],
                [127, 127, 127],
                [188, 189, 34],
                [140, 86, 75],
                [23, 190, 207],
                [227, 119, 194],
                [31, 119, 180],
                [255, 127, 14],
                [148, 103, 189],
                [255, 127, 14],
                [214, 39, 40],
                [31, 119, 180],
                [44, 160, 44],
            ],
            "keypoint_colors": [
                [148, 103, 189],
                [31, 119, 180],
                [148, 103, 189],
                [31, 119, 180],
                [148, 103, 189],
                [31, 119, 180],
                [148, 103, 189],
                [31, 119, 180],
                [148, 103, 189],
                [31, 119, 180],
                [148, 103, 189],
                [31, 119, 180],
                [148, 103, 189],
                [31, 119, 180],
                [148, 103, 189],
                [31, 119, 180],
                [148, 103, 189],
            ],
            "transforms": [
                {"KeypointsLongestMaxSize": {"max_height": 640, "max_width": 640}},
                {
                    "KeypointsPadIfNeeded": {
                        "min_height": 640,
                        "min_width": 640,
                        "image_pad_value": 127,
                        "mask_pad_value": 1,
                    }
                },
                {
                    "KeypointsRandomHorizontalFlip": {
                        "flip_index": [0, 2, 1, 4, 3, 6, 5, 8, 7, 10, 9, 12, 11, 14, 13, 16, 15],
                        "prob": 0.5,
                    }
                },
                {
                    "KeypointsRandomAffineTransform": {
                        "max_rotation": 30,
                        "min_scale": 0.5,
                        "max_scale": 2,
                        "max_translate": 0.2,
                        "image_pad_value": 127,
                        "mask_pad_value": 1,
                        "prob": 0.75,
                    }
                },
                {"KeypointsImageStandardize": {"max_value": 255}},
                {"KeypointsImageNormalize": {"mean": [0.485, 0.456, 0.406], "std": [0.229, 0.224, 0.225]}},
                "KeypointsImageToTensor",
            ],
            "target_generator": {
                "DEKRTargetsGenerator": {
                    "output_stride": 4,
                    "sigma": 2,
                    "center_sigma": 4,
                    "bg_weight": 0.1,
                    "offset_radius": 4,
                }
            },
        },
        dataloader_params={
            "shuffle": True,
            "batch_size": 24,
            "num_workers": 8,
            "drop_last": True,
            "collate_fn": "KeypointsCollate",
        },
    )

    val_dataloader = dataloaders.get(
        name="coco2017_pose_val",
        dataset_params={
            "data_dir": "/data/coco/",
            "images_dir": "images/val2017",
            "json_file": "annotations/person_keypoints_val2017.json",
            "include_empty_samples": True,
            "min_instance_area": 128,
            "edge_links": [
                [0, 1],
                [0, 2],
                [1, 2],
                [1, 3],
                [2, 4],
                [3, 5],
                [4, 6],
                [5, 6],
                [5, 7],
                [5, 11],
                [6, 8],
                [6, 12],
                [7, 9],
                [8, 10],
                [11, 12],
                [11, 13],
                [12, 14],
                [13, 15],
                [14, 16],
            ],
            "edge_colors": [
                [214, 39, 40],
                [148, 103, 189],
                [44, 160, 44],
                [140, 86, 75],
                [227, 119, 194],
                [127, 127, 127],
                [188, 189, 34],
                [127, 127, 127],
                [188, 189, 34],
                [140, 86, 75],
                [23, 190, 207],
                [227, 119, 194],
                [31, 119, 180],
                [255, 127, 14],
                [148, 103, 189],
                [255, 127, 14],
                [214, 39, 40],
                [31, 119, 180],
                [44, 160, 44],
            ],
            "keypoint_colors": [
                [148, 103, 189],
                [31, 119, 180],
                [148, 103, 189],
                [31, 119, 180],
                [148, 103, 189],
                [31, 119, 180],
                [148, 103, 189],
                [31, 119, 180],
                [148, 103, 189],
                [31, 119, 180],
                [148, 103, 189],
                [31, 119, 180],
                [148, 103, 189],
                [31, 119, 180],
                [148, 103, 189],
                [31, 119, 180],
                [148, 103, 189],
            ],
            "transforms": [
                {"KeypointsLongestMaxSize": {"max_height": 640, "max_width": 640}},
                {
                    "KeypointsPadIfNeeded": {
                        "min_height": 640,
                        "min_width": 640,
                        "image_pad_value": 127,
                        "mask_pad_value": 1,
                    }
                },
                {"KeypointsImageStandardize": {"max_value": 255}},
                {"KeypointsImageNormalize": {"mean": [0.485, 0.456, 0.406], "std": [0.229, 0.224, 0.225]}},
                "KeypointsImageToTensor",
            ],
            "target_generator": {
                "DEKRTargetsGenerator": {
                    "output_stride": 4,
                    "sigma": 2,
                    "center_sigma": 4,
                    "bg_weight": 0.1,
                    "offset_radius": 4,
                }
            },
        },
        dataloader_params={"batch_size": 32, "num_workers": 8, "drop_last": False, "collate_fn": "KeypointsCollate"},
    )

    _lr_updates = super_gradients.training.utils.utils.empty_list()

    _valid_metrics_list_0_poseestimationmetrics_post_prediction_callback = (
        super_gradients.training.utils.pose_estimation.DEKRPoseEstimationDecodeCallback(
            max_num_people=30,
            keypoint_threshold=0.05,
            nms_threshold=0.05,
            nms_num_threshold=8,
            output_stride=4,
            apply_sigmoid=True,
        )
    )

    training_hyperparams = {
        "resume": False,
        "run_id": None,
        "resume_path": None,
        "resume_from_remote_sg_logger": False,
        "ckpt_name": "ckpt_latest.pth",
        "lr_mode": "CosineLRScheduler",
        "lr_schedule_function": None,
        "lr_warmup_epochs": 0,
        "lr_warmup_steps": 256,
        "lr_cooldown_epochs": 0,
        "warmup_initial_lr": 1e-06,
        "step_lr_update_freq": None,
        "cosine_final_lr_ratio": 0.1,
        "warmup_mode": "LinearEpochLRWarmup",
        "lr_updates": _lr_updates,
        "pre_prediction_callback": None,
        "optimizer": "AdamW",
        "optimizer_params": {"weight_decay": 0.0001},
        "load_opt_params": True,
        "zero_weight_decay_on_bias_and_bn": False,
        "loss": {"DEKRLoss": {"heatmap_loss": "qfl", "heatmap_loss_factor": 1.0, "offset_loss_factor": 0.1}},
        "criterion_params": {},
        "ema": False,
        "ema_params": {"decay": 0.9997, "decay_type": "exp", "beta": 20},
        "train_metrics_list": [],
        "valid_metrics_list": [
            {
                "PoseEstimationMetrics": {
                    "num_joints": 17,
                    "oks_sigmas": [
                        0.026,
                        0.025,
                        0.025,
                        0.035,
                        0.035,
                        0.079,
                        0.079,
                        0.072,
                        0.072,
                        0.062,
                        0.062,
                        0.107,
                        0.107,
                        0.087,
                        0.087,
                        0.089,
                        0.089,
                    ],
                    "max_objects_per_image": 30,
                    "post_prediction_callback": _valid_metrics_list_0_poseestimationmetrics_post_prediction_callback,
                }
            }
        ],
        "metric_to_watch": "AP",
        "greater_metric_to_watch_is_better": True,
        "launch_tensorboard": False,
        "tensorboard_port": None,
        "tb_files_user_prompt": False,
        "save_tensorboard_to_s3": False,
        "precise_bn": False,
        "precise_bn_batch_size": None,
        "sync_bn": False,
        "silent_mode": False,
        "mixed_precision": True,
        "save_ckpt_epoch_list": [],
        "average_best_models": True,
        "dataset_statistics": False,
        "batch_accumulate": 1,
        "run_validation_freq": 1,
        "run_test_freq": 1,
        "save_model": True,
        "seed": 42,
        "phase_callbacks": [],
        "log_installed_packages": True,
        "clip_grad_norm": None,
        "ckpt_best_name": "ckpt_best.pth",
        "max_train_batches": None,
        "max_valid_batches": None,
        "sg_logger": "base_sg_logger",
        "sg_logger_params": {
            "tb_files_user_prompt": False,
            "launch_tensorboard": False,
            "tensorboard_port": None,
            "save_checkpoints_remote": False,
            "save_tensorboard_remote": False,
            "save_logs_remote": False,
            "monitor_system": True,
        },
        "torch_compile": False,
        "torch_compile_loss": False,
        "torch_compile_options": {
            "mode": "reduce-overhead",
            "fullgraph": False,
            "dynamic": False,
            "backend": "inductor",
            "options": None,
            "disable": False,
        },
        "_convert_": "all",
        "max_epochs": 150,
        "initial_lr": 0.001,
    }

    # TRAIN
    result = trainer.train(
        model=model,
        train_loader=train_dataloader,
        valid_loader=val_dataloader,
        training_params=training_hyperparams,
    )

    print(result)


if __name__ == "__main__":
    main()

Also maybe find a better name?
We already have src/super_gradients/examples/convert_recipe_example/convert_recipe_example.py

Louis-Dupont · 2023-10-29T13:59:58Z

Also maybe find a better name?
We already have src/super_gradients/examples/convert_recipe_example/convert_recipe_example.py

Maybe changing convert_from_recipe.py -> generate_script_from_recipe.py, generate_python_code_from_recipe.py or just recipe_to_code.py ?

BloodAxe · 2023-10-29T14:07:54Z

Recipe to code 🔥 🔥 🔥

# Conflicts: # tests/deci_core_unit_test_suite_runner.py

shaydeci

LGTM

BloodAxe added 6 commits October 18, 2023 10:47

Convert from recipe

69e2d67

Merge branch 'master' into feature/SG-1178-convert-from-recipe

24acf97

Merge branch 'master' into feature/SG-1178-convert-from-recipe

f33e1d0

Draft implementation of convert_from_recipe.py

4a144a7

Added global config resolve call

0ff6c31

Install black on demand

f94c9e3

Louis-Dupont reviewed Oct 24, 2023

View reviewed changes

src/super_gradients/convert_from_recipe.py Outdated Show resolved Hide resolved

shaydeci requested changes Oct 24, 2023

View reviewed changes

BloodAxe added 2 commits October 25, 2023 11:46

Added unit test to test all recipes

420aef4

Make black optional

2ede874

BloodAxe marked this pull request as ready for review October 25, 2023 10:57

BloodAxe requested a review from ofrimasad as a code owner October 25, 2023 10:57

Merge branch 'master' into feature/SG-1178-convert-from-recipe

febc43a

# Conflicts: # tests/deci_core_unit_test_suite_runner.py

shaydeci requested changes Oct 26, 2023

View reviewed changes

tests/unit_tests/convert_from_recipe_tests.py Outdated Show resolved Hide resolved

BloodAxe added 4 commits October 30, 2023 09:21

Rename script to convert_recipe_to_code

bff04ed

Fixed case when train_dataloader/val_dataloader is None

5b8089d

Merge branch 'master' into feature/SG-1178-convert-from-recipe

8719e9a

# Conflicts: # tests/deci_core_unit_test_suite_runner.py

Added step to export recipe to code and run it

da4f9f5

shaydeci previously approved these changes Oct 30, 2023

View reviewed changes

BloodAxe added 2 commits October 30, 2023 10:37

Rename import from test

5f45aaa

Rename tests to match the naming convention

17acd2c

BloodAxe dismissed shaydeci’s stale review via 17acd2c October 30, 2023 09:36

shaydeci merged commit ef8b54e into master Oct 30, 2023
7 checks passed

shaydeci deleted the feature/SG-1178-convert-from-recipe branch October 30, 2023 10:24

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Added convert_from_recipe script #1568

Added convert_from_recipe script #1568

BloodAxe commented Oct 24, 2023 •

edited

Loading

shaydeci commented Oct 26, 2023

Example usage:

How it works

Open questions

Example of the generated script file:

Louis-Dupont commented Oct 29, 2023

BloodAxe commented Oct 29, 2023

shaydeci left a comment

Added convert_from_recipe script #1568

Added convert_from_recipe script #1568

Conversation

BloodAxe commented Oct 24, 2023 • edited Loading

Example usage:

How it works

Open questions

Example of the generated script file:

shaydeci commented Oct 26, 2023

Example usage:

How it works

Open questions

Example of the generated script file:

Louis-Dupont commented Oct 29, 2023

BloodAxe commented Oct 29, 2023

shaydeci left a comment

Choose a reason for hiding this comment

BloodAxe commented Oct 24, 2023 •

edited

Loading