Skip to content

Fix activation checkpoint args for T5 #5416

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
MaximumEntropy merged 2 commits into from
Nov 15, 2022
Merged

Fix activation checkpoint args for T5 #5416

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
3afda68
Fix args (#5410)
MaximumEntropy Nov 15, 2022
6863fed
Merge branch 'main' into cherry-pick-main-1b5fac4ab0a5067df55283c1589…
MaximumEntropy Nov 15, 2022
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
1 change: 1 addition & 0 deletions examples/nlp/language_modeling/conf/megatron_model_base_config.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -31,5 +31,6 @@ onnx_safe: False # Use work-arounds for known problems with Torch ONNX exporter.
fp32_residual_connection: False # Use FP32 for residual connections.
activations_checkpoint_method: null # 'uniform', 'block'
activations_checkpoint_num_layers: 1
activations_checkpoint_granularity: null
megatron_legacy: False # Whether to use the legacy Megatron model. This affects the way q,k,v is partitioned from the mixed q,k,v layer in ParallelAttention. This needs to be True for models converted from HF.
normalize_attention_scores: True # Whether to scale the output Q * K^T by 1 / sqrt(hidden_size_per_head). This arg is provided as a configuration option mostly for compatibility with models that have been weight-converted from HF. You almost always want to se this to True.
1 change: 1 addition & 0 deletions nemo/collections/nlp/modules/common/megatron/megatron_decoders.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -119,6 +119,7 @@ def get_decoder_model(
fp32_residual_connection=fp32_residual_connection,
activations_checkpoint_method=activations_checkpoint_method,
activations_checkpoint_num_layers=activations_checkpoint_num_layers,
activations_checkpoint_granularity=activations_checkpoint_granularity,
layernorm_epsilon=layernorm_epsilon,
bias_activation_fusion=bias_activation_fusion,
bias_dropout_add_fusion=bias_dropout_add_fusion,
Expand Down
2 changes: 2 additions & 0 deletions nemo/collections/nlp/modules/common/megatron/megatron_encoders.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -121,6 +121,7 @@ def get_encoder_model(
fp32_residual_connection=fp32_residual_connection,
activations_checkpoint_method=activations_checkpoint_method,
activations_checkpoint_num_layers=activations_checkpoint_num_layers,
activations_checkpoint_granularity=activations_checkpoint_granularity,
layernorm_epsilon=layernorm_epsilon,
bias_activation_fusion=bias_activation_fusion,
bias_dropout_add_fusion=bias_dropout_add_fusion,
Expand Down Expand Up @@ -198,6 +199,7 @@ def get_encoder_model(
fp32_residual_connection=fp32_residual_connection,
activations_checkpoint_method=activations_checkpoint_method,
activations_checkpoint_num_layers=activations_checkpoint_num_layers,
activations_checkpoint_granularity=activations_checkpoint_granularity,
layernorm_epsilon=layernorm_epsilon,
bias_activation_fusion=bias_activation_fusion,
bias_dropout_add_fusion=bias_dropout_add_fusion,
Expand Down
2 changes: 2 additions & 0 deletions nemo/collections/nlp/modules/common/megatron/megatron_transformer_decoder.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -65,6 +65,7 @@ def __init__(
fp32_residual_connection=False,
activations_checkpoint_method=None,
activations_checkpoint_num_layers=1,
activations_checkpoint_granularity=None,
layernorm_epsilon=1e-5,
bias_activation_fusion=True,
bias_dropout_add_fusion=True,
Expand Down Expand Up @@ -119,6 +120,7 @@ def __init__(
fp32_residual_connection=fp32_residual_connection,
activations_checkpoint_method=activations_checkpoint_method,
activations_checkpoint_num_layers=activations_checkpoint_num_layers,
activations_checkpoint_granularity=activations_checkpoint_granularity,
layernorm_epsilon=layernorm_epsilon,
hidden_dropout=hidden_dropout,
attention_dropout=attention_dropout,
Expand Down
2 changes: 2 additions & 0 deletions nemo/collections/nlp/modules/common/megatron/megatron_transformer_encoder.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -62,6 +62,7 @@ def __init__(
fp32_residual_connection=False,
activations_checkpoint_method=None,
activations_checkpoint_num_layers=1,
activations_checkpoint_granularity=None,
layernorm_epsilon=1e-5,
bias_activation_fusion=True,
bias_dropout_add_fusion=True,
Expand Down Expand Up @@ -117,6 +118,7 @@ def __init__(
fp32_residual_connection=fp32_residual_connection,
activations_checkpoint_method=activations_checkpoint_method,
activations_checkpoint_num_layers=activations_checkpoint_num_layers,
activations_checkpoint_granularity=activations_checkpoint_granularity,
layernorm_epsilon=layernorm_epsilon,
hidden_dropout=hidden_dropout,
attention_dropout=attention_dropout,
Expand Down
4 changes: 3 additions & 1 deletion nemo/collections/nlp/modules/common/megatron/token_level_encoder_decoder.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -179,6 +179,7 @@ def __init__(
fp32_residual_connection=encoder_cfg.get('fp32_residual_connection', False),
activations_checkpoint_method=encoder_cfg.get('activations_checkpoint_method', None),
activations_checkpoint_num_layers=encoder_cfg.get('activations_checkpoint_num_layers', 1),
activations_checkpoint_granularity=encoder_cfg.get('activations_checkpoint_granularity', None),
layernorm_epsilon=encoder_cfg.get('layernorm_epsilon', 1e-5),
bias_activation_fusion=encoder_cfg.get('bias_activation_fusion', True),
bias_dropout_add_fusion=encoder_cfg.get('bias_dropout_add_fusion', True),
Expand Down Expand Up @@ -279,11 +280,12 @@ def __init__(
use_cpu_initialization=use_cpu_initialization,
hidden_dropout=decoder_cfg.get('hidden_dropout', 0.1),
attention_dropout=decoder_cfg.get('attention_dropout', 0.1),
ffn_dropout=encoder_cfg.get('ffn_dropout', 0.0),
ffn_dropout=decoder_cfg.get('ffn_dropout', 0.0),
precision=precision,
fp32_residual_connection=decoder_cfg.get('fp32_residual_connection', False),
activations_checkpoint_method=decoder_cfg.get('activations_checkpoint_method', None),
activations_checkpoint_num_layers=decoder_cfg.get('activations_checkpoint_num_layers', 1),
activations_checkpoint_granularity=decoder_cfg.get('activations_checkpoint_granularity', None),
layernorm_epsilon=decoder_cfg.get('layernorm_epsilon', 1e-5),
bias_activation_fusion=decoder_cfg.get('bias_activation_fusion', True),
bias_dropout_add_fusion=decoder_cfg.get('bias_dropout_add_fusion', True),
Expand Down