Merge branch 'master' into loadams/update-transformers-deepspeed-modules

microsoft · Dec 15, 2023 · 0efae32 · 0efae32
2 parents 3a1def0 + d1f1d45
commit 0efae32
Show file tree

Hide file tree

Showing 2 changed files with 2 additions and 2 deletions.
diff --git a/deepspeed/model_implementations/transformers/ds_transformer.py b/deepspeed/model_implementations/transformers/ds_transformer.py
@@ -163,7 +163,7 @@ def forward(
 
         if (self.config.dtype in [torch.float16, torch.bfloat16, torch.int8]) \
             and input.dtype == torch.float:
-            target_dtype = torch.half if self.dtype == torch.int8 else self.dtype
+            target_dtype = torch.half if self.config.dtype == torch.int8 else self.config.dtype
             input = input.to(target_dtype)
 
         with torch.no_grad():

diff --git a/docs/_tutorials/getting-started.md b/docs/_tutorials/getting-started.md
@@ -8,7 +8,7 @@ tags: getting-started
 ## Installation
 
 * Installing is as simple as `pip install deepspeed`, [see more details](/tutorials/advanced-install/).
-* To get started with DeepSpeed on AzureML, please see the [AzureML Examples GitHub](https://github.com/Azure/azureml-examples/tree/main/python-sdk/workflows/train/deepspeed)
+* To get started with DeepSpeed on AzureML, please see the [AzureML Examples GitHub](https://github.com/Azure/azureml-examples/tree/main/cli/jobs/deepspeed)
 * DeepSpeed has direct integrations with [HuggingFace Transformers](https://github.com/huggingface/transformers) and [PyTorch Lightning](https://github.com/PyTorchLightning/pytorch-lightning). HuggingFace Transformers users can now easily accelerate their models with DeepSpeed through a simple ``--deepspeed`` flag + config file [See more details](https://huggingface.co/docs/transformers/main_classes/deepspeed). PyTorch Lightning provides easy access to DeepSpeed through the Lightning Trainer [See more details](https://pytorch-lightning.readthedocs.io/en/stable/advanced/multi_gpu.html?highlight=deepspeed#deepspeed).
 * DeepSpeed on AMD can be used via our [ROCm images](https://hub.docker.com/r/deepspeed/rocm501/tags), e.g., `docker pull deepspeed/rocm501:ds060_pytorch110`.