activation function in BERTIntermediate #17
Merged
Conversation
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
|
Looks good, thanks for that! |
qwang70
pushed a commit
to DRL36/pytorch-pretrained-BERT
that referenced
this pull request
Mar 2, 2019
activation function in BERTIntermediate
Closed
stevezheng23
added a commit
to stevezheng23/transformers
that referenced
this pull request
Mar 24, 2020
fix at issues in roberta/berta modeling
2 tasks
jlamypoirier
added a commit
to jlamypoirier/transformers
that referenced
this pull request
Apr 4, 2023
jameshennessytempus
pushed a commit
to jameshennessytempus/transformers
that referenced
this pull request
Jun 1, 2023
1 task
jonb377
pushed a commit
to jonb377/hf-transformers
that referenced
this pull request
Nov 3, 2023
Summary: This pull request adds 2D SPMD sharding to the table. It will shard both weights and activations. Here is the sharding strategy. Let's say we have a 2D mesh (data, model) and data x model == num_devices: 1. input (data,, None, model) 2. embedding (model, data) 3. attn QKV (data, model) 4. attn O (model, data) 5. mlp gate, up (model, data) 6. mlp down (data, model) 7. activation (data,, None, model) Currently you can specify the model dimension using a new option --spmd_2d_sharding, then the data dimension will be auto-calculated. TODO: maybe we should have another option to specify whether or not we should shard the activations/inputs or shard them differently.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Was previously hardcoded to gelu because pretrained BERT models use gelu.
Changed to make BERTIntermediate use functions and "gelu", "relu" or "swish" from
config.