Skip to content

Commit 9fa7581

Browse files
nv-kkudrynskinv-kkudrynski
committed
Merge branch 'gh/master' into gh/release
2 parents 360d1f4 + 1f92262 commit 9fa7581

2 files changed

Lines changed: 2 additions & 2 deletions

File tree

  • PyTorch/LanguageModeling/BERT/triton
  • TensorFlow/Segmentation/UNet_Medical

‎PyTorch/LanguageModeling/BERT/triton/README.md‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -13,7 +13,7 @@ The first step is to train BERT for question answering. The process is the same
1313

1414
2. Build the Docker container with `bash ./scripts/docker/build.sh`.
1515

16-
3. [train](https://gitlab-master.nvidia.com/dl/JoC/bert_pyt#training-process) your own checkpoint and fine-tune it, or [download](https://ngc.nvidia.com/catalog/models/nvidia:bert_large_pyt_amp_ckpt_squad_qa1_1/files) the already trained and fine-tuned checkpoint from the [NGC](https://ngc.nvidia.com/catalog/models/nvidia:bert_large_pyt_amp_ckpt_squad_qa1_1/files) model repository.
16+
3. [train](https://github.com/NVIDIA/DeepLearningExamples/tree/master/PyTorch/LanguageModeling/BERT#training-process) your own checkpoint and fine-tune it, or [download](https://ngc.nvidia.com/catalog/models/nvidia:bert_large_pyt_amp_ckpt_squad_qa1_1/files) the already trained and fine-tuned checkpoint from the [NGC](https://ngc.nvidia.com/catalog/models/nvidia:bert_large_pyt_amp_ckpt_squad_qa1_1/files) model repository.
1717

1818
The checkpoint should be placed in `[bert folder]/checkpoints/<checkpoint>`. By default, the scripts assume `<checkpoint>` is `bert_qa.pt`, therefore, you might have to rename the trained or downloaded models as necessary.
1919

‎TensorFlow/Segmentation/UNet_Medical/README.md‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -518,7 +518,7 @@ Our results were obtained by running the `examples/unet_TRAIN_BENCHMARK{_TF-AMP}
518518
| GPUs | Batch size / GPU | Throughput - TF32 [img/s] | Throughput - mixed precision [img/s] | Throughput speedup (TF32 - mixed precision) | Weak scaling - TF32 | Weak scaling - mixed precision |
519519
|:----:|:----------------:|:-------------------------:|:------------------------------------:|:-------------------------------------------:|:-------------------:|:------------------------------:|
520520
| 1 | 1 | 29.81 | 64.22 | 2.15 | - | - |
521-
| 1 | 8 | 46.53 | 120.08 | 2.58 | - | - |
521+
| 1 | 8 | 40.50 | 120.08 | 2.58 | - | - |
522522
| 8 | 1 | 169.62 | 293.31 | 1.73 | 5.69 | 4.57 |
523523
| 8 | 8 | 304.64 | 738.64 | 2.42 | 6.55 | 6.15 |
524524

0 commit comments

Comments
 (0)