Joshua Perk
commited on
Commit
•
b13dd2b
1
Parent(s):
d66da82
Revert to deepset model only
Browse files- config.json +2 -2
- decoder_model.onnx +0 -3
- decoder_with_past_model.onnx +0 -3
- encoder_model.onnx +0 -3
- old/export_model.py → export_model.py +11 -0
- handler.py +0 -17
- old-req.txt +0 -24
- old/config.json +0 -3
- old/merges.txt +0 -0
- old/special_tokens_map.json +0 -3
- old/tokenizer.json +0 -3
- old/tokenizer_config.json +0 -3
- old/vocab.json +0 -3
- old/pytorch_model.bin → pytorch_model.bin +0 -0
- requirements.txt +24 -1
- tokenizer_config.json +2 -2
config.json
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
-
size
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:e1d8e25d7e2253e2cdf052086eb65616b5c239cfa17febbeb22c28cb9601b9c8
|
3 |
+
size 1703
|
decoder_model.onnx
DELETED
@@ -1,3 +0,0 @@
|
|
1 |
-
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:ff3b4a6bb62ddd4fe4f9b541a9c6197326118fb855ea2785bb66d1687b0d86d8
|
3 |
-
size 1222666655
|
|
|
|
|
|
decoder_with_past_model.onnx
DELETED
@@ -1,3 +0,0 @@
|
|
1 |
-
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:ea74662431a57df2675e9d950d629e42aab2fa18c929094135e7c80203f6f627
|
3 |
-
size 1121888515
|
|
|
|
|
|
encoder_model.onnx
DELETED
@@ -1,3 +0,0 @@
|
|
1 |
-
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:1b1a5e7021e95441c44514c23c4316548c4bd1e428559aea2c9359787ca67a7f
|
3 |
-
size 814841685
|
|
|
|
|
|
old/export_model.py → export_model.py
RENAMED
@@ -1,11 +1,22 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
from optimum.onnxruntime import ORTModelForSeq2SeqLM
|
2 |
from transformers import AutoTokenizer
|
|
|
3 |
model_checkpoint = "getvector/earnings-transcript-summary"
|
4 |
save_directory = "tmp/onnx/"
|
5 |
huggingface_auth_token = "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"
|
|
|
6 |
# Load a model from transformers and export it to ONNX
|
7 |
ort_model = ORTModelForSeq2SeqLM.from_pretrained(model_checkpoint, from_transformers=True, use_auth_token=huggingface_auth_token)
|
8 |
tokenizer = AutoTokenizer.from_pretrained(model_checkpoint, use_auth_token=huggingface_auth_token)
|
|
|
9 |
# Save the onnx model and tokenizer
|
10 |
ort_model.save_pretrained(save_directory)
|
11 |
tokenizer.save_pretrained(save_directory)
|
1 |
+
'''
|
2 |
+
A script to convert this normal/deepset model into an onnx model for further optimization.
|
3 |
+
Use the local .env and install the requirements
|
4 |
+
Make sure the model checkpoint name is correct (or convert to do this locally)
|
5 |
+
Make sure the auth token has access to the model
|
6 |
+
'''
|
7 |
+
|
8 |
+
|
9 |
from optimum.onnxruntime import ORTModelForSeq2SeqLM
|
10 |
from transformers import AutoTokenizer
|
11 |
+
|
12 |
model_checkpoint = "getvector/earnings-transcript-summary"
|
13 |
save_directory = "tmp/onnx/"
|
14 |
huggingface_auth_token = "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"
|
15 |
+
|
16 |
# Load a model from transformers and export it to ONNX
|
17 |
ort_model = ORTModelForSeq2SeqLM.from_pretrained(model_checkpoint, from_transformers=True, use_auth_token=huggingface_auth_token)
|
18 |
tokenizer = AutoTokenizer.from_pretrained(model_checkpoint, use_auth_token=huggingface_auth_token)
|
19 |
+
|
20 |
# Save the onnx model and tokenizer
|
21 |
ort_model.save_pretrained(save_directory)
|
22 |
tokenizer.save_pretrained(save_directory)
|
handler.py
DELETED
@@ -1,17 +0,0 @@
|
|
1 |
-
from optimum.onnxruntime import ORTModelForSeq2SeqLM
|
2 |
-
from transformers import pipeline, AutoTokenizer
|
3 |
-
|
4 |
-
class EndpointHandler():
|
5 |
-
def __init__(self, path=""):
|
6 |
-
model = ORTModelForSeq2SeqLM.from_pretrained(path)
|
7 |
-
tokenizer = AutoTokenizer.from_pretrained(path)
|
8 |
-
self.pipeline = pipeline("summarization", model=model, tokenizer=tokenizer)
|
9 |
-
|
10 |
-
def __call__(self, data):
|
11 |
-
inputs = data.pop("inputs", data)
|
12 |
-
parameters = data.pop("parameters", None)
|
13 |
-
if parameters is not None:
|
14 |
-
summary = self.pipeline(inputs, **parameters)
|
15 |
-
else:
|
16 |
-
summary = self.pipeline(inputs)
|
17 |
-
return summary
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
old-req.txt
DELETED
@@ -1,24 +0,0 @@
|
|
1 |
-
certifi==2022.9.24
|
2 |
-
charset-normalizer==2.1.1
|
3 |
-
coloredlogs==15.0.1
|
4 |
-
filelock==3.8.0
|
5 |
-
huggingface-hub==0.10.0
|
6 |
-
humanfriendly==10.0
|
7 |
-
idna==3.4
|
8 |
-
mpmath==1.2.1
|
9 |
-
numpy==1.23.3
|
10 |
-
optimum==1.4.0
|
11 |
-
packaging==21.3
|
12 |
-
protobuf==3.20.1
|
13 |
-
pyparsing==3.0.9
|
14 |
-
PyYAML==6.0
|
15 |
-
regex==2022.9.13
|
16 |
-
requests==2.28.1
|
17 |
-
sentencepiece==0.1.97
|
18 |
-
sympy==1.11.1
|
19 |
-
tokenizers==0.12.1
|
20 |
-
torch==1.12.1
|
21 |
-
tqdm==4.64.1
|
22 |
-
transformers==4.22.2
|
23 |
-
typing_extensions==4.3.0
|
24 |
-
urllib3==1.26.12
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
old/config.json
DELETED
@@ -1,3 +0,0 @@
|
|
1 |
-
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:e1d8e25d7e2253e2cdf052086eb65616b5c239cfa17febbeb22c28cb9601b9c8
|
3 |
-
size 1703
|
|
|
|
|
|
old/merges.txt
DELETED
The diff for this file is too large to render.
See raw diff
|
old/special_tokens_map.json
DELETED
@@ -1,3 +0,0 @@
|
|
1 |
-
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:06e405a36dfe4b9604f484f6a1e619af1a7f7d09e34a8555eb0b77b66318067f
|
3 |
-
size 280
|
|
|
|
|
|
old/tokenizer.json
DELETED
@@ -1,3 +0,0 @@
|
|
1 |
-
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:4dcc90e1c6bea98fde74f85a2a307a1935fe9067bd4c9d6623c520965efce58f
|
3 |
-
size 2108689
|
|
|
|
|
|
old/tokenizer_config.json
DELETED
@@ -1,3 +0,0 @@
|
|
1 |
-
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:c3d09c165ebe460bc0c338606208e725ce9e1c24df798a156d43cd854fff12db
|
3 |
-
size 375
|
|
|
|
|
|
old/vocab.json
DELETED
@@ -1,3 +0,0 @@
|
|
1 |
-
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:ed19656ea1707df69134c4af35c8ceda2cc9860bf2c3495026153a133670ab5e
|
3 |
-
size 798293
|
|
|
|
|
|
old/pytorch_model.bin → pytorch_model.bin
RENAMED
File without changes
|
requirements.txt
CHANGED
@@ -1 +1,24 @@
|
|
1 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
certifi==2022.9.24
|
2 |
+
charset-normalizer==2.1.1
|
3 |
+
coloredlogs==15.0.1
|
4 |
+
filelock==3.8.0
|
5 |
+
huggingface-hub==0.10.0
|
6 |
+
humanfriendly==10.0
|
7 |
+
idna==3.4
|
8 |
+
mpmath==1.2.1
|
9 |
+
numpy==1.23.3
|
10 |
+
optimum==1.4.0
|
11 |
+
packaging==21.3
|
12 |
+
protobuf==3.20.1
|
13 |
+
pyparsing==3.0.9
|
14 |
+
PyYAML==6.0
|
15 |
+
regex==2022.9.13
|
16 |
+
requests==2.28.1
|
17 |
+
sentencepiece==0.1.97
|
18 |
+
sympy==1.11.1
|
19 |
+
tokenizers==0.12.1
|
20 |
+
torch==1.12.1
|
21 |
+
tqdm==4.64.1
|
22 |
+
transformers==4.22.2
|
23 |
+
typing_extensions==4.3.0
|
24 |
+
urllib3==1.26.12
|
tokenizer_config.json
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
-
size
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:c3d09c165ebe460bc0c338606208e725ce9e1c24df798a156d43cd854fff12db
|
3 |
+
size 375
|