[Core] [Bugfix] Add Input Embeddings (#15428)

Signed-off-by: Andrew Sansom <andrew@protopia.ai>
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
Co-authored-by: 临景 <linjing.yx@alibaba-inc.com>
Co-authored-by: Bryce1010 <bryceyx@gmail.com>
Co-authored-by: Nan2018 <nan@protopia.ai>
Co-authored-by: Cyrus Leung <cyrus.tl.leung@gmail.com>
Co-authored-by: DarkLight1337 <tlleungac@connect.ust.hk>
This commit is contained in:
Andrew Sansom
2025-05-02 01:06:39 -07:00
committed by GitHub
co-authored by 临景 Bryce1010 Nan2018 Cyrus Leung DarkLight1337
parent 9e2de9b9e9
commit cc2a77d7f1
22 changed files with 691 additions and 113 deletions
+4 -2
View File
@@ -167,6 +167,7 @@ class MultiStepOutputProcessor(SequenceGroupOutputProcessor):
sampling_params: SamplingParams) -> None:
output_token_ids = [sample.output_token for sample in valid_samples]
output_logprobs = [sample.logprobs for sample in valid_samples]
output_embeds = [sample.output_embed for sample in valid_samples]
# Truncate to max_tokens if necessary.
remaining_tokens = sampling_params.max_tokens - (seq.get_output_len() +
@@ -190,11 +191,12 @@ class MultiStepOutputProcessor(SequenceGroupOutputProcessor):
is_prefill_sampled_token = seq.data.get_num_uncomputed_tokens() == 0
# Incrementally append tokens to the sequence, as if we had only one new
# token.
for output_token_id, output_logprob in zip(output_token_ids,
output_logprobs):
for output_token_id, output_logprob, output_embed in zip(
output_token_ids, output_logprobs, output_embeds):
seq.append_token_id(
token_id=output_token_id,
logprobs=output_logprob,
token_embed=output_embed,
)
if is_prefill_sampled_token:
+2 -1
View File
@@ -119,7 +119,8 @@ class SingleStepOutputProcessor(SequenceGroupOutputProcessor):
sample = outputs.samples[0]
seq = seq_group.first_seq
if not is_async:
seq.append_token_id(sample.output_token, sample.logprobs)
seq.append_token_id(sample.output_token, sample.logprobs,
sample.output_embed)
if sampling_params.detokenize and self.detokenizer:
new_char_count = self.detokenizer.decode_sequence_inplace(
seq, sampling_params)