2023-05-04 07:22:54 -06:00
import datetime
2023-02-14 05:02:16 -07:00
import torch
2023-06-23 04:40:46 -06:00
import os
2023-02-14 05:02:16 -07:00
from loguru import logger
from pathlib import Path
2023-07-07 06:50:12 -06:00
from safetensors . torch import save_file , load_file , _find_shared_tensors , _is_complete
from typing import List , Dict
from collections import defaultdict
2023-02-14 05:02:16 -07:00
2023-07-07 06:50:12 -06:00
def _remove_duplicate_names (
state_dict : Dict [ str , torch . Tensor ] ,
* ,
preferred_names : List [ str ] = None ,
discard_names : List [ str ] = None ,
) - > Dict [ str , List [ str ] ] :
if preferred_names is None :
preferred_names = [ ]
preferred_names = set ( preferred_names )
if discard_names is None :
discard_names = [ ]
discard_names = set ( discard_names )
shareds = _find_shared_tensors ( state_dict )
to_remove = defaultdict ( list )
for shared in shareds :
complete_names = set (
[ name for name in shared if _is_complete ( state_dict [ name ] ) ]
)
if not complete_names :
fit for baichuan models (#981)
As more and more people begin to use Baichuan's open-source models, the
influence of Baichuan models is growing, especially in China. Many
community members are interested in adding support for Baichuan models
to TGI. Meanwhile, Baichuan is a very open company, and in the future,
it plans to open-source more and more models, taking all this into
consideration, we would like to add support for the Baichuan model to
TGI. To do this, we need to make some changes, which we hope can be
merged into the main branch of TGI. In the future, we would be happy to
help maintain support for Baichuan models in TGI. We sincerely hope that
our pull request can be accepted. Thank you.
By the way, the changes of this time mainly for supporting Baichuan-7B.
---------
Co-authored-by: xiaoyuze <xiaoyuze@baichuan.com>
Co-authored-by: Nicolas Patry <patry.nicolas@protonmail.com>
2023-09-08 08:51:34 -06:00
if len ( shared ) == 1 :
# Force contiguous
name = list ( shared ) [ 0 ]
state_dict [ name ] = state_dict [ name ] . clone ( )
complete_names = { name }
else :
raise RuntimeError (
f " Error while trying to find names to remove to save state dict, but found no suitable name to keep for saving amongst: { shared } . None is covering the entire storage.Refusing to save/load the model since you could be storing much more memory than needed. Please refer to https://huggingface.co/docs/safetensors/torch_shared_tensors for more information. Or open an issue. "
)
2023-07-07 06:50:12 -06:00
keep_name = sorted ( list ( complete_names ) ) [ 0 ]
# Mecanism to preferentially select keys to keep
# coming from the on-disk file to allow
# loading models saved with a different choice
# of keep_name
preferred = complete_names . difference ( discard_names )
if preferred :
keep_name = sorted ( list ( preferred ) ) [ 0 ]
if preferred_names :
preferred = preferred_names . intersection ( complete_names )
if preferred :
keep_name = sorted ( list ( preferred ) ) [ 0 ]
for name in sorted ( shared ) :
if name != keep_name :
to_remove [ keep_name ] . append ( name )
return to_remove
def convert_file ( pt_file : Path , sf_file : Path , discard_names : List [ str ] ) :
2023-02-14 05:02:16 -07:00
"""
Convert a pytorch file to a safetensors file
2023-06-23 04:40:46 -06:00
This will remove duplicate tensors from the file .
2023-02-14 05:02:16 -07:00
2023-06-23 04:40:46 -06:00
Unfortunately , this might not respect * transformers * convention .
Forcing us to check for potentially different keys during load when looking
for specific tensors ( making tensor sharing explicit ) .
"""
loaded = torch . load ( pt_file , map_location = " cpu " )
if " state_dict " in loaded :
loaded = loaded [ " state_dict " ]
2023-07-07 06:50:12 -06:00
to_removes = _remove_duplicate_names ( loaded , discard_names = discard_names )
2023-06-23 04:40:46 -06:00
metadata = { " format " : " pt " }
for kept_name , to_remove_group in to_removes . items ( ) :
for to_remove in to_remove_group :
if to_remove not in metadata :
metadata [ to_remove ] = kept_name
del loaded [ to_remove ]
# Force tensors to be contiguous
loaded = { k : v . contiguous ( ) for k , v in loaded . items ( ) }
dirname = os . path . dirname ( sf_file )
os . makedirs ( dirname , exist_ok = True )
save_file ( loaded , sf_file , metadata = metadata )
reloaded = load_file ( sf_file )
for k in loaded :
pt_tensor = loaded [ k ]
sf_tensor = reloaded [ k ]
if not torch . equal ( pt_tensor , sf_tensor ) :
raise RuntimeError ( f " The output tensors do not match for key { k } " )
2023-02-14 05:02:16 -07:00
2023-07-07 06:50:12 -06:00
def convert_files ( pt_files : List [ Path ] , sf_files : List [ Path ] , discard_names : List [ str ] ) :
2023-05-05 09:57:02 -06:00
assert len ( pt_files ) == len ( sf_files )
2023-02-14 05:02:16 -07:00
2023-05-04 07:22:54 -06:00
N = len ( pt_files )
2023-02-14 05:02:16 -07:00
# We do this instead of using tqdm because we want to parse the logs with the launcher
2023-05-05 09:57:02 -06:00
for i , ( pt_file , sf_file ) in enumerate ( zip ( pt_files , sf_files ) ) :
2023-07-13 13:54:55 -06:00
# Skip blacklisted files
if (
" arguments " in pt_file . name
or " args " in pt_file . name
or " training " in pt_file . name
) :
continue
2023-05-05 07:28:08 -06:00
start = datetime . datetime . now ( )
2023-07-07 06:50:12 -06:00
convert_file ( pt_file , sf_file , discard_names )
2023-05-04 07:22:54 -06:00
elapsed = datetime . datetime . now ( ) - start
logger . info ( f " Convert: [ { i + 1 } / { N } ] -- Took: { elapsed } " )