Skip to content

feat: Add streaming support for batched prompts #446

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

Open
wants to merge 1 commit into
base: main
Choose a base branch
from
+5 −8
Open

feat: Add streaming support for batched prompts #446

Show file tree
Hide file tree
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
11 changes: 4 additions & 7 deletions gemma/gm/text/_sampler.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -292,11 +292,6 @@ def sample(
rng = _normalize_rng(rng)

has_batch_dim = _get_has_batch_dim(prompt)
if stream and has_batch_dim:
raise ValueError(
'Streaming is not supported for batched prompts. Let us know if you'
' need this feature.'
)

# Normalize the text, images. Tokenize, shard,...
inputs = self._get_inputs(
Expand Down Expand Up @@ -365,6 +360,7 @@ def sample(
return self._stream_decode_state( # pytype: disable=bad-return-type
state,
return_state=return_state,
has_batch_dim=has_batch_dim,
)
else:
return self._decode_state( # pytype: disable=bad-return-type
Expand Down Expand Up @@ -466,12 +462,13 @@ def _stream_decode_state(
state_iter: Iterator[_sampler_loop.SamplingState],
*,
return_state: bool,
has_batch_dim: bool,
):
for i, state in enumerate(state_iter):
yield self._decode_state(
state,
predicted_tokens=state.predicted_tokens[..., i],
has_batch_dim=False,
predicted_tokens=state.predicted_tokens[..., :i+1],
has_batch_dim=has_batch_dim,
return_state=return_state,
)

Expand Down
2 changes: 1 addition & 1 deletion gemma/gm/text/_sampler_loop.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -205,7 +205,7 @@ def _stream_sample_loop(
for _ in range(max_new_tokens):
# Exit if the cache is full.
cache = _cache_helper.Cache(state.cache)
if state.done[0].tolist() or cache.is_full:
if jnp.all(state.done) or cache.is_full:
break

state = self._sample_step(
Expand Down