-
Notifications
You must be signed in to change notification settings - Fork 2.5k
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
TE acceleration using callbacks (#11261)
* TE acceleration using callbacks Signed-off-by: Onur Yilmaz <[email protected]> * TE accelerator example added Signed-off-by: Onur Yilmaz <[email protected]> * fp8 autocast added Signed-off-by: Onur Yilmaz <[email protected]> * single GPU support with TE Signed-off-by: Onur Yilmaz <[email protected]> * generalized model callback added Signed-off-by: Onur Yilmaz <[email protected]> * Apply isort and black reformatting Signed-off-by: oyilmaz-nvidia <[email protected]> * remove te_transform Signed-off-by: Onur Yilmaz <[email protected]> * DDP with TE is working Signed-off-by: Onur Yilmaz <[email protected]> * address feedback Signed-off-by: Onur Yilmaz <[email protected]> * updated function name and added export guard Signed-off-by: Onur Yilmaz <[email protected]> * move the torch.no_grad Signed-off-by: Onur Yilmaz <[email protected]> * gemma hf example added Signed-off-by: Onur Yilmaz <[email protected]> * gemma hf example added Signed-off-by: Onur Yilmaz <[email protected]> * add docstrings Signed-off-by: Onur Yilmaz <[email protected]> * Fix the hf te test param Signed-off-by: Onur Yilmaz <[email protected]> * fix minor issue Signed-off-by: Onur Yilmaz <[email protected]> * fix the recent callback related issue Signed-off-by: Onur Yilmaz <[email protected]> --------- Signed-off-by: Onur Yilmaz <[email protected]> Signed-off-by: oyilmaz-nvidia <[email protected]> Co-authored-by: oyilmaz-nvidia <[email protected]>
- Loading branch information
1 parent
cae18e6
commit fcd128e
Showing
7 changed files
with
330 additions
and
18 deletions.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,13 @@ | ||
| # Copyright (c) 2024, NVIDIA CORPORATION. All rights reserved. | ||
| # | ||
| # Licensed under the Apache License, Version 2.0 (the "License"); | ||
| # you may not use this file except in compliance with the License. | ||
| # You may obtain a copy of the License at | ||
| # | ||
| # http://www.apache.org/licenses/LICENSE-2.0 | ||
| # | ||
| # Unless required by applicable law or agreed to in writing, software | ||
| # distributed under the License is distributed on an "AS IS" BASIS, | ||
| # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. | ||
| # See the License for the specific language governing permissions and | ||
| # limitations under the License. |
121 changes: 121 additions & 0 deletions
121
nemo/lightning/pytorch/accelerate/transformer_engine.py
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,121 @@ | ||
| # Copyright (c) 2024, NVIDIA CORPORATION. All rights reserved. | ||
| # | ||
| # Licensed under the Apache License, Version 2.0 (the "License"); | ||
| # you may not use this file except in compliance with the License. | ||
| # You may obtain a copy of the License at | ||
| # | ||
| # http://www.apache.org/licenses/LICENSE-2.0 | ||
| # | ||
| # Unless required by applicable law or agreed to in writing, software | ||
| # distributed under the License is distributed on an "AS IS" BASIS, | ||
| # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. | ||
| # See the License for the specific language governing permissions and | ||
| # limitations under the License. | ||
|
|
||
| from types import MethodType | ||
|
|
||
| import torch | ||
| from nemo.utils import logging | ||
| from nemo.utils.import_utils import safe_import_from | ||
|
|
||
| te, HAVE_TE = safe_import_from("transformer_engine", "pytorch") | ||
|
|
||
|
|
||
| def te_accelerate(model, fp8_autocast=False): | ||
| """ | ||
| Replaces original model layers with TE's accelerated layers | ||
| Args: | ||
| model: HF model | ||
| fp8_autocast (bool): apply autocast or not | ||
| """ | ||
|
|
||
| if not HAVE_TE: | ||
| logging.warning("Transformer Engine is not available and the module replacements " "will not be applied.") | ||
| else: | ||
| _apply_basic_module_replacement(model) | ||
| if fp8_autocast: | ||
| apply_fp8_autocast(model) | ||
|
|
||
|
|
||
| @torch.no_grad | ||
| def _apply_basic_module_replacement(model): | ||
| for name, module in model.named_modules(): | ||
| if isinstance(module, torch.nn.Linear): | ||
| has_bias = module.bias is not None | ||
| if any(p % 16 != 0 for p in module.weight.shape): | ||
| continue | ||
| te_module = te.Linear( | ||
| module.in_features, module.out_features, bias=has_bias, params_dtype=module.weight.dtype | ||
| ) | ||
| te_module.weight.copy_(module.weight) | ||
| if has_bias: | ||
| te_module.bias.copy_(module.bias) | ||
|
|
||
| setattr(module, name.split(".")[-1], te_module) | ||
| elif isinstance(module, torch.nn.LayerNorm): | ||
| te_module = te.LayerNorm(module.normalized_shape[0], eps=module.eps, params_dtype=module.weight.dtype) | ||
| te_module.weight.copy_(module.weight) | ||
| te_module.bias.copy_(module.bias) | ||
| setattr(module, name.split(".")[-1], te_module) | ||
| elif isinstance(module, torch.nn.RMSNorm): | ||
| te_module = te.RMSNorm(module.normalized_shape[0], eps=module.eps, dtype=module.weight.dtype) | ||
| te_module.weight.copy_(module.weight) | ||
| te_module.bias.copy_(module.bias) | ||
| setattr(module, name.split(".")[-1], te_module) | ||
|
|
||
|
|
||
| def is_te_accelerated(model): | ||
| """ | ||
| Checks whether model has TE layers or not | ||
| Args: | ||
| model: HF model | ||
| """ | ||
|
|
||
| if not HAVE_TE: | ||
| logging.warning("Transformer Engine is not available.") | ||
| return False | ||
| else: | ||
| for name, module in model.named_modules(): | ||
| if isinstance(module, (te.LayerNorm, te.Linear, te.TransformerLayer)): | ||
| return True | ||
|
|
||
| return False | ||
|
|
||
|
|
||
| def apply_fp8_autocast(model, fp8_recipe_handler=None): | ||
| """ | ||
| Applies TE's autocast | ||
| Args: | ||
| model: HF model | ||
| fp8_recipe_handler: fpt handler | ||
| """ | ||
|
|
||
| if not HAVE_TE: | ||
| logging.warning("Transformer Engine is not available and the FP8 autocast " "will not be applied.") | ||
| else: | ||
| import transformer_engine.common.recipe as te_recipe | ||
|
|
||
| kwargs = fp8_recipe_handler.to_kwargs() if fp8_recipe_handler is not None else {} | ||
| if "fp8_format" in kwargs: | ||
| kwargs["fp8_format"] = getattr(te_recipe.Format, kwargs["fp8_format"]) | ||
| use_during_eval = kwargs.pop("use_autocast_during_eval", False) | ||
| fp8_recipe = te_recipe.DelayedScaling(**kwargs) | ||
| new_forward = _contextual_fp8_autocast(model.forward, fp8_recipe, use_during_eval) | ||
|
|
||
| if hasattr(model.forward, "__func__"): | ||
| model.forward = MethodType(new_forward, model) | ||
| else: | ||
| model.forward = new_forward | ||
|
|
||
|
|
||
| def _contextual_fp8_autocast(model_forward, fp8_recipe, use_during_eval=False): | ||
| from transformer_engine.pytorch import fp8_autocast | ||
|
|
||
| def forward(self, *args, **kwargs): | ||
| enabled = use_during_eval or self.training | ||
| with fp8_autocast(enabled=enabled, fp8_recipe=fp8_recipe): | ||
| return model_forward(*args, **kwargs) | ||
|
|
||
| forward.__wrapped__ = model_forward | ||
|
|
||
| return forward |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Oops, something went wrong.