huggingface/text-generation-inference
5
0
Fork 0
mirror of https://github.com/huggingface/text-generation-inference.git synced 2025-09-09 19:34:53 +00:00
Code Issues Packages Projects Releases Wiki Activity

cleanup

Browse Source
This commit is contained in:
OlivierDehaene 2023-05-10 18:20:28 +02:00
parent 3248fdfbd4
commit a944dd0fd5
1 changed files with 13 additions and 13 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

26
server/text_generation_server/utils/tokens.py
View File

@ -29,7 +29,7 @@ class Sampling:
q = torch.empty_like(probs).exponential_(1, generator=self.generator)
torch.div(probs, q, out=q)
return torch.argmax(q, dim=-1, keepdim=True)
return q.argmax()
class Greedy:
@ -107,36 +107,36 @@ class NextTokenChooser:
seed=0,
device="cpu",
):
self.watermark_warper = (
self.watermark_processor = (
WatermarkLogitsProcessor(device=device) if watermark else None
)
self.repetition_warper = (
self.repetition_processor = (
RepetitionPenaltyLogitsProcessor(penalty=repetition_penalty)
if repetition_penalty
else None
)
sampling = (
do_sample
or (temperature is not None and temperature != 1.0)
has_warpers = (
(temperature is not None and temperature != 1.0)
or (top_k is not None and top_k != 0)
or (top_p is not None and top_p < 1.0)
or (typical_p is not None and typical_p < 1.0)
)
if sampling:
self.choice = Sampling(seed, device)
if has_warpers:
self.static_warper = static_warper(
temperature=temperature, top_k=top_k, top_p=top_p, typical_p=typical_p
)
else:
self.choice = Greedy()
self.static_warper = None
sampling = do_sample or has_warpers
self.choice = Sampling(seed, device) if sampling else Greedy()
def __call__(self, input_ids, scores):
if self.watermark_warper:
scores = self.watermark_warper(input_ids, scores)
if self.repetition_warper:
scores = self.repetition_warper(input_ids, scores)
if self.watermark_processor:
scores = self.watermark_processor(input_ids, scores)
if self.repetition_processor:
scores = self.repetition_processor(input_ids, scores)
if self.static_warper is None:
next_logprob = torch.log_softmax(scores, -1)