Commit 9b0083ea authored by Vijay Korthikanti's avatar Vijay Korthikanti
Browse files

Incrementing checkpoint version to 2.0

parent ee327acd
...@@ -101,7 +101,7 @@ def save_checkpoint(iteration, model, optimizer, lr_scheduler): ...@@ -101,7 +101,7 @@ def save_checkpoint(iteration, model, optimizer, lr_scheduler):
# Arguments, iteration, and model. # Arguments, iteration, and model.
state_dict = {} state_dict = {}
state_dict['args'] = args state_dict['args'] = args
state_dict['checkpoint_version'] = 1.0 state_dict['checkpoint_version'] = 2.0
state_dict['iteration'] = iteration state_dict['iteration'] = iteration
state_dict['model'] = model.state_dict_for_save_checkpoint() state_dict['model'] = model.state_dict_for_save_checkpoint()
......
...@@ -172,19 +172,28 @@ class ParallelSelfAttention(MegatronModule): ...@@ -172,19 +172,28 @@ class ParallelSelfAttention(MegatronModule):
init_method=output_layer_init_method, init_method=output_layer_init_method,
skip_bias_add=True) skip_bias_add=True)
def _transpose_last_dim(self, mixed_layer, num_splits): def _transpose_last_dim(self, mixed_layer, num_splits, num_splits_index):
"""[s, b, num_splits * np * hn] """[s, b, num_splits * np * hn]
-->(view) [s, b, num_splits, np, hn] -->(view) [s, b, num_splits, np, hn]
-->(tranpose) [s, b, np, num_splits, hn] -->(tranpose) [s, b, np, num_splits, hn]
-->(view) [s, b, np * num_splits * hn] """ -->(view) [s, b, np * num_splits * hn] """
input_shape = mixed_layer.size(); input_shape = mixed_layer.size();
intermediate_shape = input_shape[:-1] +\ if num_splits_index == 0:
(num_splits, self.num_attention_heads_per_partition, intermediate_shape = input_shape[:-1] +\
self.hidden_size_per_attention_head) (num_splits, self.num_attention_heads_per_partition,
self.hidden_size_per_attention_head)
mixed_layer = mixed_layer.view(*intermediate_shape) mixed_layer = mixed_layer.view(*intermediate_shape)
mixed_layer = mixed_layer.transpose(-2, -3).contiguous() mixed_layer = mixed_layer.transpose(-2, -3).contiguous()
else:
assert num_splits_index == 2
intermediate_shape = input_shape[:-1] +\
(self.num_attention_heads_per_partition,
self.hidden_size_per_attention_head, num_splits)
mixed_layer = mixed_layer.view(*intermediate_shape)
mixed_layer = mixed_layer.transpose(-1, -2).contiguous()
mixed_layer = mixed_layer.view(*input_shape) mixed_layer = mixed_layer.view(*input_shape)
return mixed_layer return mixed_layer
...@@ -201,10 +210,13 @@ class ParallelSelfAttention(MegatronModule): ...@@ -201,10 +210,13 @@ class ParallelSelfAttention(MegatronModule):
mixed_x_layer, _ = self.query_key_value(hidden_states) mixed_x_layer, _ = self.query_key_value(hidden_states)
checkpoint_version = get_checkpoint_version() checkpoint_version = get_checkpoint_version()
if checkpoint_version is not None and \ if checkpoint_version is not None:
checkpoint_version == 0: if checkpoint_version == 0:
# [s, b, (3 * np * hn)] --> [s, b, (np * 3 * hn)] # [s, b, (3 * np * hn)] --> [s, b, (np * 3 * hn)]
mixed_x_layer = self._transpose_last_dim(mixed_x_layer, 3) mixed_x_layer = self._transpose_last_dim(mixed_x_layer, 3, 0)
elif checkpoint_version == 1:
# [s, b, (np * hn * 3)] --> [s, b, (np * 3 * hn)]
mixed_x_layer = self._transpose_last_dim(mixed_x_layer, 3, 2)
# [sq, b, (np * 3 * hn)] --> [sq, b, np, 3 * hn] # [sq, b, (np * 3 * hn)] --> [sq, b, np, 3 * hn]
new_tensor_shape = mixed_x_layer.size()[:-1] + \ new_tensor_shape = mixed_x_layer.size()[:-1] + \
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment