Fix to prefix tuning to fit transformers#2096

Merged

BenjaminBossan merged 31 commits intohuggingface:mainfrom BenjaminBossan:fix-prefix-tuning-dynamic-cache

Oct 24, 2024

+71-8

Commits on Sep 25, 2024

[WIP] Fix to prefix tuning to fit transformers
BenjaminBossan
committed

Commits on Oct 22, 2024

Update src/peft/peft_model.py

BenjaminBossan
and
zucchini-nlp
authored
FIX: Change check if past_key_values is empty (huggingface#2106 )
BenjaminBossan
committed
DOC Update source install instruction (huggingface#2110 )

Salehbigdeli
authored and
BenjaminBossan
committed
FIX Refactor OFT, small changes to BOFT (huggingface#1996 )

Zeju1997
authored and
BenjaminBossan
committed
ENH: Improved attribute access for modules_to_save (huggingface#2117 )
BenjaminBossan
committed
FIX low_cpu_mem_usage consolidates devices (huggingface#2113 )
BenjaminBossan
committed
TST Mark flaky X-LoRA test as xfail (huggingface#2114 )
BenjaminBossan
committed
ENH: Warn when from_pretrained misses PEFT keys (huggingface#2118 )
BenjaminBossan
committed
FEAT: Adding exclude modules param(huggingface#2044 ) (huggingface#2102 )

JINO-ROHIT
authored and
BenjaminBossan
committed
FIX BC breaking change to boft conv2d scaling variable (huggingface#2127 )

Zeju1997
authored and
BenjaminBossan
committed
FEAT: VeRA quantization using bitsandbytes (huggingface#2070 ) (huggingface#2076 )

ZiadHelal
authored and
BenjaminBossan
committed
Bump version to 0.13.2.dev0 (huggingface#2137 )
BenjaminBossan
committed
FEAT: Support torchao (huggingface#2062 )

BenjaminBossan
and
SunMarc
committed
FIX: PiSSA now works with Conv1D layers (huggingface#2103 ) (huggingface#2104 )

authored and
BenjaminBossan
committed
FIX Type annoations in vera/bnb.py (huggingface#2139 )
BenjaminBossan
committed
ENH Make PEFT configs forward compatible (huggingface#2038 )
BenjaminBossan
committed
FIX Raise mixed adapter infer with missing adapter (huggingface#2090 )
BenjaminBossan
committed
FIX Prompt learning with latest transformers error (huggingface#2140 )
BenjaminBossan
committed
ENH LoRA notebook for NER task (huggingface#2126 )

JINO-ROHIT
authored and
BenjaminBossan
committed
FIX TST NaN issue with HQQ GPU test (huggingface#2143 )
BenjaminBossan
committed
FIX Bug in target module optimization if suffix (huggingface#2144 )
BenjaminBossan
committed
Bump version to 0.13.2.dev0 (huggingface#2145 )
BenjaminBossan
committed
FIX Don't assume past_key_valus for encoder models (huggingface#2149 )
BenjaminBossan
committed
FIX Use SFTConfig instead of SFTTrainer keyword args (huggingface#2150 )

qgallouedec
authored and
BenjaminBossan
committed
make style
BenjaminBossan
committed
Merge branch 'main' into fix-prefix-tuning-dynamic-cache
BenjaminBossan
committed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Fix to prefix tuning to fit transformers#2096

Commits on Sep 25, 2024

[WIP] Fix to prefix tuning to fit transformers

Commits on Oct 22, 2024

Update src/peft/peft_model.py

FIX: Change check if past_key_values is empty (huggingface#2106 )

DOC Update source install instruction (huggingface#2110 )

FIX Refactor OFT, small changes to BOFT (huggingface#1996 )

ENH: Improved attribute access for modules_to_save (huggingface#2117 )

FIX low_cpu_mem_usage consolidates devices (huggingface#2113 )

TST Mark flaky X-LoRA test as xfail (huggingface#2114 )

ENH: Warn when from_pretrained misses PEFT keys (huggingface#2118 )

FEAT: Adding exclude modules param(huggingface#2044 ) (huggingface#2102 )

FIX BC breaking change to boft conv2d scaling variable (huggingface#2127 )

FEAT: VeRA quantization using bitsandbytes (huggingface#2070 ) (huggingface#2076 )

Bump version to 0.13.2.dev0 (huggingface#2137 )

FEAT: Support torchao (huggingface#2062 )

FIX: PiSSA now works with Conv1D layers (huggingface#2103 ) (huggingface#2104 )

FIX Type annoations in vera/bnb.py (huggingface#2139 )

ENH Make PEFT configs forward compatible (huggingface#2038 )

FIX Raise mixed adapter infer with missing adapter (huggingface#2090 )

FIX Prompt learning with latest transformers error (huggingface#2140 )

ENH LoRA notebook for NER task (huggingface#2126 )

FIX TST NaN issue with HQQ GPU test (huggingface#2143 )

FIX Bug in target module optimization if suffix (huggingface#2144 )

Bump version to 0.13.2.dev0 (huggingface#2145 )

FIX Don't assume past_key_valus for encoder models (huggingface#2149 )

FIX Use `SFTConfig` instead of `SFTTrainer` keyword args (huggingface#2150 )

make style

Merge branch 'main' into fix-prefix-tuning-dynamic-cache

Commits on Oct 23, 2024

Some further fixes

Fixes for seq2seq model

Mistral test requires protobuf (???)

Also sentencepiece for mistral -_-