Hugging Face's logo

sergiopaniego
/
OptimizedPilotNet

TF-Keras
TF Lite
Model card Files Community
sergiopaniego commited on
Commit
d4f715f
•
1 Parent(s): 222d844

Updated TensorRT PyTorch models

Browse files
Files changed (4) hide show
  1. README.md +4 -4
  2. pytorch/{28_04_trt_mod_float_28_04_int8.jit.pt → trained_pilotNet_qat_int8.jit.pt} +0 -0
  3. pytorch/{28_04_trt_mod_float_28_04_float.jit.pt → trt_mod_float16.jit.pt} +0 -0
  4. pytorch/{28_04_trt_mod_float_28_04_half.jit.pt → trt_mod_float32.jit.pt} +0 -0
README.md CHANGED
@@ -63,7 +63,7 @@ To do inference:
63
  | Original | 6.1217 | 0.03524 | - | 28_04_pilot_net_model_best_123.pth |
64
  | Dynamic Range Quantization | 1.9493608474731445 | 0.012065857842182075 | 0.001480283498764038 | 24_05_dynamic_quan.pth |
65
  | Static Quantization | 1.6071176528930664 | 0.012072610909984047 | 0.0007314345836639404 | 24_05_static_quan.pth |
66
- | Quantization Aware Training | 1.6069536209106445 | 0.01109830549109022 | 0.0011710402965545653 | 24_05_quan_aware.pth |
67
  | Local Prune | 6.122584342956543 | 0.010850968803449539 | 0.0014387350082397461 | 24_05_local_prune.pth |
68
  | Global Prune | 6.122775077819824 | 0.010964057565769462 | 0.0014179635047912597 | 24_05_global_prune.pth |
69
  | Prune + Quantization | 1.6067094802856445 | 0.010949893930274941 | 0.0011728739738464356 | 24_05_prune_quan.pth |
@@ -78,9 +78,9 @@ To do inference:
78
 
79
  | Optimization | Model size (MB) | MSE | Inference time (s/frame) | Filename |
80
  | --------------------------------- | ---------------------- | ---------------------- | ---------------------- | --------------------------------- |
81
- | Float32 Quantization | 6.121363639831543 | 0.009570527376262128 | 0.0002284455299377441 | 28_04_trt_mod_float_28_04_float.jit.pt |
82
- | Float16 Quantization | 6.121363639831543 | 0.009571507916721152 | 0.000250823974609375 | 28_04_trt_mod_float_28_04_half.jit.pt |
83
- | Int8 Quantization | 6.181861877441406 | 0.00969304293365875 | 0.0002463934421539307 | 28_04_trt_mod_float_28_04_int8.jit.pt |
84
 
85
 
86
 
 
63
  | Original | 6.1217 | 0.03524 | - | 28_04_pilot_net_model_best_123.pth |
64
  | Dynamic Range Quantization | 1.9493608474731445 | 0.012065857842182075 | 0.001480283498764038 | 24_05_dynamic_quan.pth |
65
  | Static Quantization | 1.6071176528930664 | 0.012072610909984047 | 0.0007314345836639404 | 24_05_static_quan.pth |
66
+ | Quantization Aware Training | 1.6069536209106445 | 0.01109830549109022 | 0.0011710402965545653 | ls .pth |
67
  | Local Prune | 6.122584342956543 | 0.010850968803449539 | 0.0014387350082397461 | 24_05_local_prune.pth |
68
  | Global Prune | 6.122775077819824 | 0.010964057565769462 | 0.0014179635047912597 | 24_05_global_prune.pth |
69
  | Prune + Quantization | 1.6067094802856445 | 0.010949893930274941 | 0.0011728739738464356 | 24_05_prune_quan.pth |
 
78
 
79
  | Optimization | Model size (MB) | MSE | Inference time (s/frame) | Filename |
80
  | --------------------------------- | ---------------------- | ---------------------- | ---------------------- | --------------------------------- |
81
+ | Float32 Quantization | 6.121363639831543 | 0.009570527376262128 | 0.0002284455299377441 | trt_mod_float32.jit.pt |
82
+ | Float16 Quantization | 6.121363639831543 | 0.009571507916721152 | 0.000250823974609375 | trt_mod_float16.jit.pt |
83
+ | Int8 Quantization | 6.181861877441406 | 0.00969304293365875 | 0.0002463934421539307 | trained_pilotNet_qat_int8.jit.pt |
84
 
85
 
86
 
pytorch/{28_04_trt_mod_float_28_04_int8.jit.pt → trained_pilotNet_qat_int8.jit.pt} RENAMED
File without changes
pytorch/{28_04_trt_mod_float_28_04_float.jit.pt → trt_mod_float16.jit.pt} RENAMED
File without changes
pytorch/{28_04_trt_mod_float_28_04_half.jit.pt → trt_mod_float32.jit.pt} RENAMED
File without changes