Skip to content

Fix tabular data text generation #7031

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
merged 1 commit into from
Jul 14, 2023
Merged

Fix tabular data text generation #7031

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
6 changes: 4 additions & 2 deletions nemo/collections/nlp/modules/common/text_generation_utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -396,6 +396,7 @@ def synced_generate(
context_length_tensor,
tokens_to_generate,
all_probs,
compute_attention_mask=compute_attention_mask,
temperature=temperature,
)
else:
Expand Down Expand Up @@ -825,6 +826,7 @@ def tab_sample_sequence_batch(
context_lengths,
tokens_to_generate,
all_probs=True,
compute_attention_mask=True,
type_ids=None,
temperature=None,
):
Expand All @@ -848,7 +850,7 @@ def tab_sample_sequence_batch(
# initialize the batch
with torch.no_grad():
context_length = context_lengths.min().item()
inference_strategy.init_batch(context_tokens, context_length)
inference_strategy.init_batch(context_tokens, context_length, compute_attention_mask)
context = context_tokens[:, :context_length]
# the context may start in the middle of the row,
# calculate the offset according to the position of '\n' or '<|endoftext|>'
Expand Down Expand Up @@ -882,7 +884,7 @@ def tab_sample_sequence_batch(

while context_length < maxlen:
batch, tensor_shape = inference_strategy.prepare_batch_at_step(
tokens, maxlen, micro_batch_size, counter, context_length
tokens, maxlen, micro_batch_size, counter, context_length, compute_attention_mask
)
output = inference_strategy.forward_step(batch, tensor_shape)

Expand Down