Просмотр исходного кода

Document behaviour when --num_iter < --warmup_steps

Lukasz Pierscieniewski 5 лет назад
Родитель
Сommit
4f8aaa22b0

+ 2 - 0
TensorFlow/Classification/ConvNets/resnet50v1.5/README.md

@@ -464,6 +464,7 @@ Each of these scripts runs 200 warm-up iterations and measures the first epoch.
 
 
 To control warmup and benchmark length, use the `--warmup_steps`, `--num_iter` and `--iter_unit` flags. Features like XLA or DALI can be controlled
 To control warmup and benchmark length, use the `--warmup_steps`, `--num_iter` and `--iter_unit` flags. Features like XLA or DALI can be controlled
 with `--use_xla` and `--use_dali` flags. If no `--data_dir=<path to imagenet>` flag is specified then the benchmarks will use a synthetic dataset.
 with `--use_xla` and `--use_dali` flags. If no `--data_dir=<path to imagenet>` flag is specified then the benchmarks will use a synthetic dataset.
+For proper throughput reporting the value of `--num_iter` must be greater than `--warmup_steps` value.
 Suggested batch sizes for training are 256 for mixed precision training and 128 for single precision training per single V100 16 GB.
 Suggested batch sizes for training are 256 for mixed precision training and 128 for single precision training per single V100 16 GB.
 
 
 #### Inference performance benchmark
 #### Inference performance benchmark
@@ -480,6 +481,7 @@ To benchmark the inference performance on a specific batch size, run:
 
 
 By default, each of these scripts runs 20 warm-up iterations and measures the next 80 iterations.
 By default, each of these scripts runs 20 warm-up iterations and measures the next 80 iterations.
 To control warm-up and benchmark length, use the `--warmup_steps`, `--num_iter` and `--iter_unit` flags. 
 To control warm-up and benchmark length, use the `--warmup_steps`, `--num_iter` and `--iter_unit` flags. 
+For proper throughput and latency reporting the value of `--num_iter` must be greater than `--warmup_steps` value.
 If no `--data_dir=<path to imagenet>` flag is specified then the benchmarks will use a synthetic dataset.
 If no `--data_dir=<path to imagenet>` flag is specified then the benchmarks will use a synthetic dataset.
 
 
 The benchmark can be automated with the `inference_benchmark.sh` script provided in `resnet50v1.5`, by simply running:
 The benchmark can be automated with the `inference_benchmark.sh` script provided in `resnet50v1.5`, by simply running:

+ 2 - 0
TensorFlow/Classification/ConvNets/resnext101-32x4d/README.md

@@ -430,6 +430,7 @@ Each of these scripts runs 200 warm-up iterations and measures the first epoch.
 
 
 To control warmup and benchmark length, use the `--warmup_steps`, `--num_iter` and `--iter_unit` flags. Features like XLA or DALI can be controlled
 To control warmup and benchmark length, use the `--warmup_steps`, `--num_iter` and `--iter_unit` flags. Features like XLA or DALI can be controlled
 with `--use_xla` and `--use_dali` flags. If no `--data_dir=<path to imagenet>` flag is specified then the benchmarks will use a synthetic dataset.
 with `--use_xla` and `--use_dali` flags. If no `--data_dir=<path to imagenet>` flag is specified then the benchmarks will use a synthetic dataset.
+For proper throughput reporting the value of `--num_iter` must be greater than `--warmup_steps` value.
 Suggested batch sizes for training are 128 for mixed precision training and 64 for single precision training per single V100 16 GB.
 Suggested batch sizes for training are 128 for mixed precision training and 64 for single precision training per single V100 16 GB.
 
 
 
 
@@ -447,6 +448,7 @@ To benchmark the inference performance on a specific batch size, run:
 
 
 By default, each of these scripts runs 20 warm-up iterations and measures the next 80 iterations.
 By default, each of these scripts runs 20 warm-up iterations and measures the next 80 iterations.
 To control warm-up and benchmark length, use the `--warmup_steps`, `--num_iter` and `--iter_unit` flags.
 To control warm-up and benchmark length, use the `--warmup_steps`, `--num_iter` and `--iter_unit` flags.
+For proper throughput and latency reporting the value of `--num_iter` must be greater than `--warmup_steps` value.
 If no `--data_dir=<path to imagenet>` flag is specified then the benchmarks will use a synthetic dataset.
 If no `--data_dir=<path to imagenet>` flag is specified then the benchmarks will use a synthetic dataset.
 
 
 The benchmark can be automated with the `inference_benchmark.sh` script provided in `resnext101-32x4d`, by simply running:
 The benchmark can be automated with the `inference_benchmark.sh` script provided in `resnext101-32x4d`, by simply running:

+ 2 - 0
TensorFlow/Classification/ConvNets/se-resnext101-32x4d/README.md

@@ -425,6 +425,7 @@ Each of these scripts runs 200 warm-up iterations and measures the first epoch.
 
 
 To control warmup and benchmark length, use the `--warmup_steps`, `--num_iter` and `--iter_unit` flags. Features like XLA or DALI can be controlled
 To control warmup and benchmark length, use the `--warmup_steps`, `--num_iter` and `--iter_unit` flags. Features like XLA or DALI can be controlled
 with `--use_xla` and `--use_dali` flags. If no `--data_dir=<path to imagenet>` flag is specified then the benchmarks will use a synthetic dataset.
 with `--use_xla` and `--use_dali` flags. If no `--data_dir=<path to imagenet>` flag is specified then the benchmarks will use a synthetic dataset.
+For proper throughput reporting the value of `--num_iter` must be greater than `--warmup_steps` value.
 Suggested batch sizes for training are 96 for mixed precision training and 64 for single precision training per single V100 16 GB.
 Suggested batch sizes for training are 96 for mixed precision training and 64 for single precision training per single V100 16 GB.
 
 
 
 
@@ -442,6 +443,7 @@ To benchmark the inference performance on a specific batch size, run:
 
 
 By default, each of these scripts runs 20 warm-up iterations and measures the next 80 iterations.
 By default, each of these scripts runs 20 warm-up iterations and measures the next 80 iterations.
 To control warm-up and benchmark length, use the `--warmup_steps`, `--num_iter` and `--iter_unit` flags.
 To control warm-up and benchmark length, use the `--warmup_steps`, `--num_iter` and `--iter_unit` flags.
+For proper throughput and latency reporting the value of `--num_iter` must be greater than `--warmup_steps` value.
 If no `--data_dir=<path to imagenet>` flag is specified then the benchmarks will use a synthetic dataset.
 If no `--data_dir=<path to imagenet>` flag is specified then the benchmarks will use a synthetic dataset.
 
 
 The benchmark can be automated with the `inference_benchmark.sh` script provided in `se-resnext101-32x4d`, by simply running:
 The benchmark can be automated with the `inference_benchmark.sh` script provided in `se-resnext101-32x4d`, by simply running: