Skip to content

[FlaxSpeechEncoderDecoder] Fix dtype bug #16581

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
patrickvonplaten merged 2 commits into from
Apr 4, 2022
Merged

[FlaxSpeechEncoderDecoder] Fix dtype bug #16581

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
777e803
[FlaxSpeechEncoderDecoder] Fix dtype bug
patrickvonplaten Apr 4, 2022
a729923
more fixes
patrickvonplaten Apr 4, 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
14 changes: 7 additions & 7 deletions src/transformers/models/speech_encoder_decoder/modeling_flax_speech_encoder_decoder.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -310,7 +310,7 @@ def __call__(
decoder_hidden_states=decoder_outputs.hidden_states,
decoder_attentions=decoder_outputs.attentions,
cross_attentions=decoder_outputs.cross_attentions,
encoder_last_hidden_state=encoder_outputs.last_hidden_state,
encoder_last_hidden_state=encoder_hidden_states,
encoder_hidden_states=encoder_outputs.hidden_states,
encoder_attentions=encoder_outputs.attentions,
)
Expand Down Expand Up @@ -363,8 +363,8 @@ def init_weights(self, rng: jax.random.PRNGKey, input_shape: Tuple) -> FrozenDic
encoder_input_shape, decoder_input_shape = input_shape

# init input DeviceArrays
inputs = jnp.zeros(encoder_input_shape, dtype="i4")
attention_mask = jnp.ones_like(inputs)
inputs = jnp.zeros(encoder_input_shape, dtype="f4")
attention_mask = jnp.ones_like(inputs, dtype="i4")
decoder_input_ids = jnp.zeros(decoder_input_shape, dtype="i4")
decoder_attention_mask = jnp.ones_like(decoder_input_ids)

Expand Down Expand Up @@ -472,7 +472,7 @@ def encode(
return_dict = return_dict if return_dict is not None else self.config.return_dict

if attention_mask is None:
attention_mask = jnp.ones_like(inputs)
attention_mask = jnp.ones_like(inputs, dtype="i4")

# Handle any PRNG if needed
rngs = {}
Expand All @@ -485,7 +485,7 @@ def _encoder_forward(module, inputs, attention_mask, **kwargs):

outputs = self.module.apply(
{"params": params or self.params},
inputs=jnp.array(inputs, dtype="i4"),
inputs=jnp.array(inputs, dtype="f4"),
attention_mask=jnp.array(attention_mask, dtype="i4"),
output_attentions=output_attentions,
output_hidden_states=output_hidden_states,
Expand Down Expand Up @@ -680,7 +680,7 @@ def __call__(

# prepare encoder inputs
if attention_mask is None:
attention_mask = jnp.ones_like(inputs)
attention_mask = jnp.ones_like(inputs, dtype="i4")

# prepare decoder inputs
if decoder_input_ids is None:
Expand All @@ -700,7 +700,7 @@ def __call__(

return self.module.apply(
{"params": params or self.params},
inputs=jnp.array(inputs, dtype="i4"),
inputs=jnp.array(inputs, dtype="f4"),
Copy link
Contributor Author

@patrickvonplaten patrickvonplaten Apr 4, 2022

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@sanchit-gandhi - this is the reason for why the script fails. We should not convert the input to int here, we're dealing with float numbers

attention_mask=jnp.array(attention_mask, dtype="i4"),
decoder_input_ids=jnp.array(decoder_input_ids, dtype="i4"),
decoder_attention_mask=jnp.array(decoder_attention_mask, dtype="i4"),
Expand Down