Victor Lavrenko PRO

lavrenko's activity

• All
• Models
• Datasets
• Spaces
• Buckets
• Papers
• Collections
• Community
• Posts
• Upvotes
• Likes
• Articles

New activity in transformers-community/group-beam-search 2 months ago

Optimize group beam search by sharing the initial prefill across beams

3

#6 opened 2 months ago by

lavrenko

Possible performance issue: group beam search appears to prefill identical prefixes once per beam

3

#5 opened 3 months ago by

lavrenko

New activity in transformers-community/sink_cache 3 months ago

"sink_cache" fails on Transformers 5.12 due to outdated "Cache.__init__" API

1

#1 opened 3 months ago by

lavrenko

New activity in transformers-community/dola 3 months ago

Fix Transformers v5 cached decoding in DoLa

2

#2 opened 3 months ago by

lavrenko

New activity in transformers-community/constrained-beam-search 3 months ago

constrained-beam-search produces degenerate output with use_cache=True

3

#2 opened 3 months ago by

lavrenko

Support Transformers v5 cache handling

2

#3 opened 3 months ago by

lavrenko

New activity in transformers-community/dola 3 months ago

DoLa custom generation produces gibberish with use_cache=True

1

#1 opened 3 months ago by

lavrenko

New activity in transformers-community/group-beam-search 3 months ago

Custom group-beam-search changes deterministic output when cache is enabled

3

#3 opened 3 months ago by

lavrenko

Support Transformers v5 cache handling

2

#4 opened 3 months ago by

lavrenko

Support Transformers v5 cache handling

2

#4 opened 3 months ago by

lavrenko

Custom group-beam-search changes deterministic output when cache is enabled

3

#3 opened 3 months ago by

lavrenko