Skip to content

Commit 8bd6dd1

Browse files
hXl3shXl3s
committed
Document synthetic dataset options
1 parent 446c878 commit 8bd6dd1

3 files changed

Lines changed: 9 additions & 7 deletions

File tree

‎TensorFlow/Classification/ConvNets/resnet50v1.5/README.md‎

Lines changed: 3 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -194,7 +194,7 @@ To train your model using mixed precision or TF32 with Tensor Cores or FP32, per
194194
1. Clone the repository.
195195
```
196196
git clone https://github.com/NVIDIA/DeepLearningExamples
197-
cd DeepLearningExamples/TensorFlow/Classification/RN50v1.5
197+
cd DeepLearningExamples/TensorFlow/Classification/ConvNets
198198
```
199199

200200
2. Download and preprocess the dataset.
@@ -452,10 +452,9 @@ To benchmark the training performance on a specific batch size, run:
452452
Each of these scripts runs 200 warm-up iterations and measures the first epoch.
453453

454454
To control warmup and benchmark length, use the `--warmup_steps`, `--num_iter` and `--iter_unit` flags. Features like XLA or DALI can be controlled
455-
with `--use_xla` and `--use_dali` flags.
455+
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.
456456
Suggested batch sizes for training are 256 for mixed precision training and 128 for single precision training per single V100 16 GB.
457457

458-
459458
#### Inference performance benchmark
460459

461460
To benchmark the inference performance on a specific batch size, run:
@@ -470,6 +469,7 @@ To benchmark the inference performance on a specific batch size, run:
470469

471470
By default, each of these scripts runs 20 warm-up iterations and measures the next 80 iterations.
472471
To control warm-up and benchmark length, use the `--warmup_steps`, `--num_iter` and `--iter_unit` flags.
472+
If no `--data_dir=<path to imagenet>` flag is specified then the benchmarks will use a synthetic dataset.
473473

474474
The benchmark can be automated with the `inference_benchmark.sh` script provided in `resnet50v1.5`, by simply running:
475475
`bash ./resnet50v1.5/inference_benchmark.sh <data dir> <data idx dir>`

‎TensorFlow/Classification/ConvNets/resnext101-32x4d/README.md‎

Lines changed: 3 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -203,7 +203,7 @@ To train your model using mixed precision or TF32 with Tensor Cores or FP32, per
203203
1. Clone the repository.
204204
```
205205
git clone https://github.com/NVIDIA/DeepLearningExamples
206-
cd DeepLearningExamples/TensorFlow/Classification/RN50v1.5
206+
cd DeepLearningExamples/TensorFlow/Classification/ConvNets
207207
```
208208

209209
2. Download and preprocess the dataset.
@@ -420,7 +420,7 @@ To benchmark the training performance on a specific batch size, run:
420420
Each of these scripts runs 200 warm-up iterations and measures the first epoch.
421421

422422
To control warmup and benchmark length, use the `--warmup_steps`, `--num_iter` and `--iter_unit` flags. Features like XLA or DALI can be controlled
423-
with `--use_xla` and `--use_dali` flags.
423+
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.
424424
Suggested batch sizes for training are 128 for mixed precision training and 64 for single precision training per single V100 16 GB.
425425

426426

@@ -438,6 +438,7 @@ To benchmark the inference performance on a specific batch size, run:
438438

439439
By default, each of these scripts runs 20 warm-up iterations and measures the next 80 iterations.
440440
To control warm-up and benchmark length, use the `--warmup_steps`, `--num_iter` and `--iter_unit` flags.
441+
If no `--data_dir=<path to imagenet>` flag is specified then the benchmarks will use a synthetic dataset.
441442

442443
The benchmark can be automated with the `inference_benchmark.sh` script provided in `resnext101-32x4d`, by simply running:
443444
`bash ./resnext101-32x4d/inference_benchmark.sh <data dir> <data idx dir>`

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

Lines changed: 3 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -198,7 +198,7 @@ To train your model using mixed precision or TF32 with Tensor Cores or FP32, per
198198
1. Clone the repository.
199199
```
200200
git clone https://github.com/NVIDIA/DeepLearningExamples
201-
cd DeepLearningExamples/TensorFlow/Classification/RN50v1.5
201+
cd DeepLearningExamples/TensorFlow/Classification/ConvNets
202202
```
203203

204204
2. Download and preprocess the dataset.
@@ -415,7 +415,7 @@ To benchmark the training performance on a specific batch size, run:
415415
Each of these scripts runs 200 warm-up iterations and measures the first epoch.
416416

417417
To control warmup and benchmark length, use the `--warmup_steps`, `--num_iter` and `--iter_unit` flags. Features like XLA or DALI can be controlled
418-
with `--use_xla` and `--use_dali` flags.
418+
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.
419419
Suggested batch sizes for training are 96 for mixed precision training and 64 for single precision training per single V100 16 GB.
420420

421421

@@ -433,6 +433,7 @@ To benchmark the inference performance on a specific batch size, run:
433433

434434
By default, each of these scripts runs 20 warm-up iterations and measures the next 80 iterations.
435435
To control warm-up and benchmark length, use the `--warmup_steps`, `--num_iter` and `--iter_unit` flags.
436+
If no `--data_dir=<path to imagenet>` flag is specified then the benchmarks will use a synthetic dataset.
436437

437438
The benchmark can be automated with the `inference_benchmark.sh` script provided in `se-resnext101-32x4d`, by simply running:
438439
`bash ./se-resnext101-32x4d/inference_benchmark.sh <data dir> <data idx dir>`

0 commit comments

Comments
 (0)