saner mask creation? (it doesnt matter, kv cache wont work)

This commit is contained in:
mrq 2024-11-02 21:00:21 -05:00
parent ded746e157
commit 3826f9bae4
3 changed files with 16 additions and 9 deletions

View File

@ -513,6 +513,8 @@ def get_task_symmap():
}
def _replace_file_extension(path, suffix):
if not isinstance( path, Path ):
path = Path(path)
return (path.parent / path.name.split(".")[0]).with_suffix(suffix)
def _get_quant_extension():

View File

@ -72,7 +72,7 @@ def process(
# easy way to load the model and handle encoding audio
if tts is None:
tts = init_tts( yaml=yaml, restart=False, device=device, dtype=dtype )
tts = init_tts( config=yaml, restart=False, device=device, dtype=dtype )
features = { key: None for key in metadata_keys }

View File

@ -78,9 +78,11 @@ def list_to_tensor(x_list: list[Tensor], pattern="t b c -> b t c"):
l = list(map(len, x_list))
x = rearrange(pad_sequence(x_list), pattern)
m = _create_mask(l, x_list[0].device)
"""
m = m.t().unsqueeze(-1) # (t b 1)
m = rearrange(m, pattern)
m = m.to(x)
"""
m = m.to(x).int()
return x, m
def _interleave_sequence_reshape( input: list[torch.Tensor], dim=-1 ):
@ -835,7 +837,7 @@ class Base(nn.Module):
output_hidden_states = False,
):
x = inputs
m = mask.squeeze(-1).int()
m = mask #.squeeze(-1).int()
aux_loss = None
attentions = None
@ -844,7 +846,7 @@ class Base(nn.Module):
# HF transformer derived model
if self.arch_type in ["llama", "mistral", "mixtral"]:
kwargs = dict(
attention_mask=m,
#attention_mask=m,
inputs_embeds=x,
past_key_values=state,
position_ids=position_ids,
@ -1475,7 +1477,9 @@ class Base(nn.Module):
return metrics["logits_entropy"] < kwargs["logits_entropy"] and metrics["logits_varentropy"] < kwargs["logits_varentropy"]
x_list = self.inputs_to_embeddings( inputs, quant_levels )
x, m = list_to_tensor(x_list)
x, mask = list_to_tensor(x_list)
m = mask.unsqueeze(dim=-1)
training = self.training
device = x.device
@ -1501,16 +1505,17 @@ class Base(nn.Module):
# pad mask
shape[2] = 1
padding = torch.zeros(shape, dtype=x.dtype, device=x.device)
m = torch.cat([m, padding], dim=1)
mask = torch.cat([mask, padding], dim=1)
# needs to be done here as we still have our raw inputs
position_ids = self.inputs_to_position_ids( inputs, mask=m.squeeze(-1).int() ) if not self.unified_position_ids else None
#position_ids = self.inputs_to_position_ids( inputs, mask=m.squeeze(-1).int() ) if not self.unified_position_ids else None
position_ids = self.inputs_to_position_ids( inputs, mask=mask ) if not self.unified_position_ids else None
classifier_quant_levels = [ -1 if inputs[i][0][-1] in self.special_tasks else l for i, l in enumerate( quant_levels ) ]
output = self._forward(
inputs=x,
mask=m,
mask=mask,
state=state,
position_ids=position_ids,
output_attentions = output_attentions,
@ -1530,7 +1535,7 @@ class Base(nn.Module):
hidden_states[i] = self.classifier(hidden_states[i]) * m
# to-do: piece-wise classification, now that there's a head for text
# although again, one single monolithic head would be preferable instead......
if self.classifiers is not None:
elif self.classifiers is not None:
logits = self.classifiers(logits, levels = classifier_quant_levels) * m
if hidden_states is not None: