Register
Login
Resources
Docs Blog Datasets Glossary Case Studies Tutorials & Webinars
Product
Data Engine LLMs Platform Enterprise
Pricing Explore
Connect to our Discord channel
Resources  
Docs Blog Datasets Glossary Case Studies Tutorials & Webinars
Explore Pricing
Product  
Data Engine LLMs Platform Enterprise
Connect to Dagshub Discord channel
Register
Login

Deci-AI
/
super-gradients
connected to https://github.com/Deci-AI/super-gradients.git

Watch 4
Issues 62 Pull Requests 44 Discussions 0

#643 PPYolo-E

Merged
Ghost merged 1 commits into Deci-AI:master from deci-ai:feature/SG-344-PP-Yolo-E-Training-Replicate-Recipe
Conversation 0 Files changed 46 Commits 1
File Compare
super-gradients
/
src
/
super_gradients
/
recipes
/
training_hyperparams
/
coco2017_ppyoloe_train_params.yaml
View File
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60 
  1. defaults:
    2. 

  2.   - default_train_params
    3. 

  3. 

  4. max_epochs: 500
    5. 

  5. static_assigner_end_epoch: 150
    6. 

  6. 

  7. warmup_mode: "linear_batch_step"
    8. 

  8. warmup_initial_lr:  1e-6
    9. 

  9. lr_warmup_steps: 1000
    10. 

  10. lr_warmup_epochs: 0
    11. 

  11. 

  12. initial_lr:  2e-3
    13. 

  13. lr_mode: cosine
    14. 

  14. cosine_final_lr_ratio: 0.1
    15. 

  15. 

  16. zero_weight_decay_on_bias_and_bn: False
    17. 

  17. batch_accumulate: 1
    18. 

  18. 

  19. save_ckpt_epoch_list: [200, 250, 300, 350, 400, 450]
    20. 

  20. 

  21. loss:
    22. 

  22.   ppyoloe_loss:
    23. 

  23.     num_classes: ${arch_params.num_classes}
    24. 

  24.     reg_max: ${arch_params.head.reg_max}
    25. 

  25. 

  26. optimizer: AdamW
    27. 

  27. optimizer_params:
    28. 

  28.   weight_decay: 0.0001
    29. 

  29. 

  30. ema: True
    31. 

  31. ema_params:
    32. 

  32.   decay: 0.9997
    33. 

  33.   decay_type: threshold
    34. 

  34. 

  35. mixed_precision: False
    36. 

  36. sync_bn: True
    37. 

  37. 

  38. valid_metrics_list:
    39. 

  39.   - DetectionMetrics:
    40. 

  40.       score_thres: 0.1
    41. 

  41.       top_k_predictions: 300
    42. 

  42.       num_cls: ${arch_params.num_classes}
    43. 

  43.       normalize_targets: True
    44. 

  44.       post_prediction_callback:
    45. 

  45.         _target_: super_gradients.training.models.detection_models.pp_yolo_e.PPYoloEPostPredictionCallback
    46. 

  46.         score_threshold: 0.01
    47. 

  47.         nms_top_k: 1000
    48. 

  48.         max_predictions: 300
    49. 

  49.         nms_threshold: 0.7
    50. 

  50. 

  51. pre_prediction_callback:
    52. 

  52. 

  53. phase_callbacks:
    54. 

  54.   - PPYoloETrainingStageSwitchCallback:
    55. 

  55.       static_assigner_end_epoch: ${training_hyperparams.static_assigner_end_epoch}
    56. 

  56. 

  57. metric_to_watch: 'mAP@0.50:0.95'
    58. 

  58. greater_metric_to_watch_is_better: True
    59. 

  59. 

  60. _convert_: all
    61.
Discard
Prev
Next
Tip!

Press p or to see the previous file or, n or to see the next file