offlinedemo 0.2.0__tar.gz → 0.4.0__tar.gz
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- offlinedemo-0.4.0/PKG-INFO +109 -0
- offlinedemo-0.4.0/README.md +42 -0
- offlinedemo-0.4.0/pyproject.toml +77 -0
- offlinedemo-0.4.0/pyproject.toml.orig +76 -0
- offlinedemo-0.2.0/PKG-INFO +0 -12
- offlinedemo-0.2.0/README.md +0 -3
- offlinedemo-0.2.0/pyproject.toml +0 -18
- offlinedemo-0.2.0/pyproject.toml.orig +0 -17
- {offlinedemo-0.2.0 → offlinedemo-0.4.0}/src/offlinedemo/__init__.py +0 -0
|
@@ -0,0 +1,109 @@
|
|
|
1
|
+
Metadata-Version: 2.3
|
|
2
|
+
Name: offlinedemo
|
|
3
|
+
Version: 0.4.0
|
|
4
|
+
Summary: Easy demonstration of models by offlineisbetter.
|
|
5
|
+
Author: offlineisbetter
|
|
6
|
+
Author-email: offlineisbetter <dev@offlineisbetter.com>
|
|
7
|
+
Requires-Dist: annotated-doc==0.0.5
|
|
8
|
+
Requires-Dist: anyio==4.15.1
|
|
9
|
+
Requires-Dist: certifi==2026.7.22
|
|
10
|
+
Requires-Dist: click==8.5.0
|
|
11
|
+
Requires-Dist: cuda-bindings==13.4.3
|
|
12
|
+
Requires-Dist: cuda-pathfinder==1.8.2
|
|
13
|
+
Requires-Dist: cuda-toolkit==13.0.3.0
|
|
14
|
+
Requires-Dist: filelock==4.0.6
|
|
15
|
+
Requires-Dist: flatbuffers==25.12.19
|
|
16
|
+
Requires-Dist: fsspec==2026.9.0
|
|
17
|
+
Requires-Dist: h11==0.16.0
|
|
18
|
+
Requires-Dist: hf-xet==1.6.0
|
|
19
|
+
Requires-Dist: httpcore==1.0.9
|
|
20
|
+
Requires-Dist: httpx==0.28.1
|
|
21
|
+
Requires-Dist: huggingface-hub==1.33.0
|
|
22
|
+
Requires-Dist: idna==3.20
|
|
23
|
+
Requires-Dist: jinja2==3.1.6
|
|
24
|
+
Requires-Dist: markdown-it-py==4.2.0
|
|
25
|
+
Requires-Dist: markupsafe==3.0.3
|
|
26
|
+
Requires-Dist: mdurl==0.1.2
|
|
27
|
+
Requires-Dist: ml-dtypes==0.6.0
|
|
28
|
+
Requires-Dist: mpmath==1.3.0
|
|
29
|
+
Requires-Dist: networkx==3.7
|
|
30
|
+
Requires-Dist: numpy==2.5.3
|
|
31
|
+
Requires-Dist: nvidia-cublas==13.1.1.3
|
|
32
|
+
Requires-Dist: nvidia-cuda-cupti==13.0.85
|
|
33
|
+
Requires-Dist: nvidia-cuda-nvrtc==13.0.88
|
|
34
|
+
Requires-Dist: nvidia-cuda-runtime==13.0.96
|
|
35
|
+
Requires-Dist: nvidia-cudnn-cu13==9.24.0.43
|
|
36
|
+
Requires-Dist: nvidia-cufft==12.0.0.61
|
|
37
|
+
Requires-Dist: nvidia-cufile==1.15.1.6
|
|
38
|
+
Requires-Dist: nvidia-curand==10.4.0.35
|
|
39
|
+
Requires-Dist: nvidia-cusolver==12.0.4.66
|
|
40
|
+
Requires-Dist: nvidia-cusparse==12.6.3.3
|
|
41
|
+
Requires-Dist: nvidia-cusparselt-cu13==0.8.1
|
|
42
|
+
Requires-Dist: nvidia-nccl-cu13==2.30.7
|
|
43
|
+
Requires-Dist: nvidia-nvjitlink==13.4.92
|
|
44
|
+
Requires-Dist: nvidia-nvshmem-cu13==3.4.5
|
|
45
|
+
Requires-Dist: nvidia-nvtx==13.0.85
|
|
46
|
+
Requires-Dist: onnx==1.23.1
|
|
47
|
+
Requires-Dist: onnxruntime==1.30.0
|
|
48
|
+
Requires-Dist: packaging==26.3
|
|
49
|
+
Requires-Dist: protobuf==7.36.2
|
|
50
|
+
Requires-Dist: pygments==2.21.0
|
|
51
|
+
Requires-Dist: pyyaml==6.0.3
|
|
52
|
+
Requires-Dist: regex==2026.9.29
|
|
53
|
+
Requires-Dist: rich==15.0.0
|
|
54
|
+
Requires-Dist: safetensors==0.8.0
|
|
55
|
+
Requires-Dist: setuptools==84.0.0
|
|
56
|
+
Requires-Dist: shellingham==1.5.4
|
|
57
|
+
Requires-Dist: sympy==1.14.0
|
|
58
|
+
Requires-Dist: tokenizers==0.23.2
|
|
59
|
+
Requires-Dist: torch==2.14.0
|
|
60
|
+
Requires-Dist: tqdm==4.70.1
|
|
61
|
+
Requires-Dist: transformers==5.17.0
|
|
62
|
+
Requires-Dist: triton==3.8.0
|
|
63
|
+
Requires-Dist: typer==0.27.2
|
|
64
|
+
Requires-Dist: typing-extensions==4.16.0
|
|
65
|
+
Requires-Python: >=3.13
|
|
66
|
+
Description-Content-Type: text/markdown
|
|
67
|
+
|
|
68
|
+
# _offlineisbetter_
|
|
69
|
+
|
|
70
|
+
inference for models by _offlineisbetter_.
|
|
71
|
+
|
|
72
|
+
we believe that you shouldn't give your data to faceless companies, that you deserve to run text models locally, and that you shouldn't need to buy expensive hardware. so we're building _offlineisbetter_.
|
|
73
|
+
|
|
74
|
+
_offlineisbetter_ models will be for _encoding_ tasks: sentiment analysis, text tagging, document retrieval, etc., rather than for _decoding_ tasks like autoregressive generation. we believe that it's wasteful and dangerous to depend on cloud apis for frontier language models to do these simple tasks, and it should be almost mindless to download a model to _use_ it without dealing with runtimes or quantization formats.
|
|
75
|
+
|
|
76
|
+
## try it yourself
|
|
77
|
+
|
|
78
|
+
try our first model yourself. our first model is a small (230m) text model for sentiment analysis called `offline-sentiment-small`.
|
|
79
|
+
|
|
80
|
+
```bash
|
|
81
|
+
pip install offlinedemo
|
|
82
|
+
```
|
|
83
|
+
|
|
84
|
+
download the model archive from the website and unpack the model.
|
|
85
|
+
|
|
86
|
+
```bash
|
|
87
|
+
tar -xvf offline-sentiment-small.tar
|
|
88
|
+
```
|
|
89
|
+
|
|
90
|
+
run the demo, passing the inflated directory containing the model checkpoint.
|
|
91
|
+
|
|
92
|
+
```bash
|
|
93
|
+
offlinedemo offline-sentiment-small
|
|
94
|
+
```
|
|
95
|
+
|
|
96
|
+
## benchmarks
|
|
97
|
+
|
|
98
|
+
below are benchmarks for `offline-sentiment-small` on binary sentiment classification using the [stanfordnlp/sst2](https://huggingface.co/datasets/stanfordnlp/sst2) validation set. all benchmarks were completed on the ryzen 9950x3d cpu on one thread.
|
|
99
|
+
|
|
100
|
+
| model | parameters | p95 (ms) | f1 (validation) |
|
|
101
|
+
|:---:|:---:|:---:|:---:|
|
|
102
|
+
| `offline-sentiment-small` | 230m | 80.32 | 0.9489 |
|
|
103
|
+
| `distilbert-base` | 67m | 66.15 | 0.9321 |
|
|
104
|
+
| `roberta-base` | 125m | 469.65 | 0.9396 |
|
|
105
|
+
| `modernbert-base` | 149m | 530.73 | 0.9396 |
|
|
106
|
+
|
|
107
|
+
## philosophy
|
|
108
|
+
|
|
109
|
+
succinctly, the core philosophy of _offlineisbetter_ is that parameter-efficient and low-latency models should be easily accessible to everybody. of course hugging face and `transformers.pipeline` allows you to run sentiment analysis in three lines of python, but for more parameter-efficient models, already quantized and with optimized computation graphs.
|
|
@@ -0,0 +1,42 @@
|
|
|
1
|
+
# _offlineisbetter_
|
|
2
|
+
|
|
3
|
+
inference for models by _offlineisbetter_.
|
|
4
|
+
|
|
5
|
+
we believe that you shouldn't give your data to faceless companies, that you deserve to run text models locally, and that you shouldn't need to buy expensive hardware. so we're building _offlineisbetter_.
|
|
6
|
+
|
|
7
|
+
_offlineisbetter_ models will be for _encoding_ tasks: sentiment analysis, text tagging, document retrieval, etc., rather than for _decoding_ tasks like autoregressive generation. we believe that it's wasteful and dangerous to depend on cloud apis for frontier language models to do these simple tasks, and it should be almost mindless to download a model to _use_ it without dealing with runtimes or quantization formats.
|
|
8
|
+
|
|
9
|
+
## try it yourself
|
|
10
|
+
|
|
11
|
+
try our first model yourself. our first model is a small (230m) text model for sentiment analysis called `offline-sentiment-small`.
|
|
12
|
+
|
|
13
|
+
```bash
|
|
14
|
+
pip install offlinedemo
|
|
15
|
+
```
|
|
16
|
+
|
|
17
|
+
download the model archive from the website and unpack the model.
|
|
18
|
+
|
|
19
|
+
```bash
|
|
20
|
+
tar -xvf offline-sentiment-small.tar
|
|
21
|
+
```
|
|
22
|
+
|
|
23
|
+
run the demo, passing the inflated directory containing the model checkpoint.
|
|
24
|
+
|
|
25
|
+
```bash
|
|
26
|
+
offlinedemo offline-sentiment-small
|
|
27
|
+
```
|
|
28
|
+
|
|
29
|
+
## benchmarks
|
|
30
|
+
|
|
31
|
+
below are benchmarks for `offline-sentiment-small` on binary sentiment classification using the [stanfordnlp/sst2](https://huggingface.co/datasets/stanfordnlp/sst2) validation set. all benchmarks were completed on the ryzen 9950x3d cpu on one thread.
|
|
32
|
+
|
|
33
|
+
| model | parameters | p95 (ms) | f1 (validation) |
|
|
34
|
+
|:---:|:---:|:---:|:---:|
|
|
35
|
+
| `offline-sentiment-small` | 230m | 80.32 | 0.9489 |
|
|
36
|
+
| `distilbert-base` | 67m | 66.15 | 0.9321 |
|
|
37
|
+
| `roberta-base` | 125m | 469.65 | 0.9396 |
|
|
38
|
+
| `modernbert-base` | 149m | 530.73 | 0.9396 |
|
|
39
|
+
|
|
40
|
+
## philosophy
|
|
41
|
+
|
|
42
|
+
succinctly, the core philosophy of _offlineisbetter_ is that parameter-efficient and low-latency models should be easily accessible to everybody. of course hugging face and `transformers.pipeline` allows you to run sentiment analysis in three lines of python, but for more parameter-efficient models, already quantized and with optimized computation graphs.
|
|
@@ -0,0 +1,77 @@
|
|
|
1
|
+
[project]
|
|
2
|
+
name = "offlinedemo"
|
|
3
|
+
version = "0.4.0"
|
|
4
|
+
description = "Easy demonstration of models by offlineisbetter."
|
|
5
|
+
readme = "README.md"
|
|
6
|
+
requires-python = ">=3.13"
|
|
7
|
+
dependencies = [
|
|
8
|
+
"annotated-doc==0.0.5",
|
|
9
|
+
"anyio==4.15.1",
|
|
10
|
+
"certifi==2026.7.22",
|
|
11
|
+
"click==8.5.0",
|
|
12
|
+
"cuda-bindings==13.4.3",
|
|
13
|
+
"cuda-pathfinder==1.8.2",
|
|
14
|
+
"cuda-toolkit==13.0.3.0",
|
|
15
|
+
"filelock==4.0.6",
|
|
16
|
+
"flatbuffers==25.12.19",
|
|
17
|
+
"fsspec==2026.9.0",
|
|
18
|
+
"h11==0.16.0",
|
|
19
|
+
"hf-xet==1.6.0",
|
|
20
|
+
"httpcore==1.0.9",
|
|
21
|
+
"httpx==0.28.1",
|
|
22
|
+
"huggingface-hub==1.33.0",
|
|
23
|
+
"idna==3.20",
|
|
24
|
+
"jinja2==3.1.6",
|
|
25
|
+
"markdown-it-py==4.2.0",
|
|
26
|
+
"markupsafe==3.0.3",
|
|
27
|
+
"mdurl==0.1.2",
|
|
28
|
+
"ml-dtypes==0.6.0",
|
|
29
|
+
"mpmath==1.3.0",
|
|
30
|
+
"networkx==3.7",
|
|
31
|
+
"numpy==2.5.3",
|
|
32
|
+
"nvidia-cublas==13.1.1.3",
|
|
33
|
+
"nvidia-cuda-cupti==13.0.85",
|
|
34
|
+
"nvidia-cuda-nvrtc==13.0.88",
|
|
35
|
+
"nvidia-cuda-runtime==13.0.96",
|
|
36
|
+
"nvidia-cudnn-cu13==9.24.0.43",
|
|
37
|
+
"nvidia-cufft==12.0.0.61",
|
|
38
|
+
"nvidia-cufile==1.15.1.6",
|
|
39
|
+
"nvidia-curand==10.4.0.35",
|
|
40
|
+
"nvidia-cusolver==12.0.4.66",
|
|
41
|
+
"nvidia-cusparse==12.6.3.3",
|
|
42
|
+
"nvidia-cusparselt-cu13==0.8.1",
|
|
43
|
+
"nvidia-nccl-cu13==2.30.7",
|
|
44
|
+
"nvidia-nvjitlink==13.4.92",
|
|
45
|
+
"nvidia-nvshmem-cu13==3.4.5",
|
|
46
|
+
"nvidia-nvtx==13.0.85",
|
|
47
|
+
"onnx==1.23.1",
|
|
48
|
+
"onnxruntime==1.30.0",
|
|
49
|
+
"packaging==26.3",
|
|
50
|
+
"protobuf==7.36.2",
|
|
51
|
+
"pygments==2.21.0",
|
|
52
|
+
"pyyaml==6.0.3",
|
|
53
|
+
"regex==2026.9.29",
|
|
54
|
+
"rich==15.0.0",
|
|
55
|
+
"safetensors==0.8.0",
|
|
56
|
+
"setuptools==84.0.0",
|
|
57
|
+
"shellingham==1.5.4",
|
|
58
|
+
"sympy==1.14.0",
|
|
59
|
+
"tokenizers==0.23.2",
|
|
60
|
+
"torch==2.14.0",
|
|
61
|
+
"tqdm==4.70.1",
|
|
62
|
+
"transformers==5.17.0",
|
|
63
|
+
"triton==3.8.0",
|
|
64
|
+
"typer==0.27.2",
|
|
65
|
+
"typing-extensions==4.16.0",
|
|
66
|
+
]
|
|
67
|
+
|
|
68
|
+
[[project.authors]]
|
|
69
|
+
name = "offlineisbetter"
|
|
70
|
+
email = "dev@offlineisbetter.com"
|
|
71
|
+
|
|
72
|
+
[project.scripts]
|
|
73
|
+
offlinedemo = "offlinedemo:main"
|
|
74
|
+
|
|
75
|
+
[build-system]
|
|
76
|
+
requires = ["uv_build>=0.12.19,<0.13.0"]
|
|
77
|
+
build-backend = "uv_build"
|
|
@@ -0,0 +1,76 @@
|
|
|
1
|
+
[project]
|
|
2
|
+
name = "offlinedemo"
|
|
3
|
+
version = "0.4.0"
|
|
4
|
+
description = "Easy demonstration of models by offlineisbetter."
|
|
5
|
+
readme = "README.md"
|
|
6
|
+
authors = [
|
|
7
|
+
{ name = "offlineisbetter", email = "dev@offlineisbetter.com" }
|
|
8
|
+
]
|
|
9
|
+
requires-python = ">=3.13"
|
|
10
|
+
dependencies = [
|
|
11
|
+
"annotated-doc==0.0.5",
|
|
12
|
+
"anyio==4.15.1",
|
|
13
|
+
"certifi==2026.7.22",
|
|
14
|
+
"click==8.5.0",
|
|
15
|
+
"cuda-bindings==13.4.3",
|
|
16
|
+
"cuda-pathfinder==1.8.2",
|
|
17
|
+
"cuda-toolkit==13.0.3.0",
|
|
18
|
+
"filelock==4.0.6",
|
|
19
|
+
"flatbuffers==25.12.19",
|
|
20
|
+
"fsspec==2026.9.0",
|
|
21
|
+
"h11==0.16.0",
|
|
22
|
+
"hf-xet==1.6.0",
|
|
23
|
+
"httpcore==1.0.9",
|
|
24
|
+
"httpx==0.28.1",
|
|
25
|
+
"huggingface-hub==1.33.0",
|
|
26
|
+
"idna==3.20",
|
|
27
|
+
"jinja2==3.1.6",
|
|
28
|
+
"markdown-it-py==4.2.0",
|
|
29
|
+
"markupsafe==3.0.3",
|
|
30
|
+
"mdurl==0.1.2",
|
|
31
|
+
"ml-dtypes==0.6.0",
|
|
32
|
+
"mpmath==1.3.0",
|
|
33
|
+
"networkx==3.7",
|
|
34
|
+
"numpy==2.5.3",
|
|
35
|
+
"nvidia-cublas==13.1.1.3",
|
|
36
|
+
"nvidia-cuda-cupti==13.0.85",
|
|
37
|
+
"nvidia-cuda-nvrtc==13.0.88",
|
|
38
|
+
"nvidia-cuda-runtime==13.0.96",
|
|
39
|
+
"nvidia-cudnn-cu13==9.24.0.43",
|
|
40
|
+
"nvidia-cufft==12.0.0.61",
|
|
41
|
+
"nvidia-cufile==1.15.1.6",
|
|
42
|
+
"nvidia-curand==10.4.0.35",
|
|
43
|
+
"nvidia-cusolver==12.0.4.66",
|
|
44
|
+
"nvidia-cusparse==12.6.3.3",
|
|
45
|
+
"nvidia-cusparselt-cu13==0.8.1",
|
|
46
|
+
"nvidia-nccl-cu13==2.30.7",
|
|
47
|
+
"nvidia-nvjitlink==13.4.92",
|
|
48
|
+
"nvidia-nvshmem-cu13==3.4.5",
|
|
49
|
+
"nvidia-nvtx==13.0.85",
|
|
50
|
+
"onnx==1.23.1",
|
|
51
|
+
"onnxruntime==1.30.0",
|
|
52
|
+
"packaging==26.3",
|
|
53
|
+
"protobuf==7.36.2",
|
|
54
|
+
"pygments==2.21.0",
|
|
55
|
+
"pyyaml==6.0.3",
|
|
56
|
+
"regex==2026.9.29",
|
|
57
|
+
"rich==15.0.0",
|
|
58
|
+
"safetensors==0.8.0",
|
|
59
|
+
"setuptools==84.0.0",
|
|
60
|
+
"shellingham==1.5.4",
|
|
61
|
+
"sympy==1.14.0",
|
|
62
|
+
"tokenizers==0.23.2",
|
|
63
|
+
"torch==2.14.0",
|
|
64
|
+
"tqdm==4.70.1",
|
|
65
|
+
"transformers==5.17.0",
|
|
66
|
+
"triton==3.8.0",
|
|
67
|
+
"typer==0.27.2",
|
|
68
|
+
"typing-extensions==4.16.0",
|
|
69
|
+
]
|
|
70
|
+
|
|
71
|
+
[project.scripts]
|
|
72
|
+
offlinedemo = "offlinedemo:main"
|
|
73
|
+
|
|
74
|
+
[build-system]
|
|
75
|
+
requires = ["uv_build>=0.12.19,<0.13.0"]
|
|
76
|
+
build-backend = "uv_build"
|
offlinedemo-0.2.0/PKG-INFO
DELETED
|
@@ -1,12 +0,0 @@
|
|
|
1
|
-
Metadata-Version: 2.3
|
|
2
|
-
Name: offlinedemo
|
|
3
|
-
Version: 0.2.0
|
|
4
|
-
Summary: Easy demonstration of models by offlineisbetter.
|
|
5
|
-
Author: offlineisbetter
|
|
6
|
-
Author-email: offlineisbetter <dev@offlineisbetter.com>
|
|
7
|
-
Requires-Python: >=3.13
|
|
8
|
-
Description-Content-Type: text/markdown
|
|
9
|
-
|
|
10
|
-
# _offlineisbetter_
|
|
11
|
-
|
|
12
|
-
Inference for _offlineisbetter_ models.
|
offlinedemo-0.2.0/README.md
DELETED
offlinedemo-0.2.0/pyproject.toml
DELETED
|
@@ -1,18 +0,0 @@
|
|
|
1
|
-
[project]
|
|
2
|
-
name = "offlinedemo"
|
|
3
|
-
version = "0.2.0"
|
|
4
|
-
description = "Easy demonstration of models by offlineisbetter."
|
|
5
|
-
readme = "README.md"
|
|
6
|
-
requires-python = ">=3.13"
|
|
7
|
-
dependencies = []
|
|
8
|
-
|
|
9
|
-
[[project.authors]]
|
|
10
|
-
name = "offlineisbetter"
|
|
11
|
-
email = "dev@offlineisbetter.com"
|
|
12
|
-
|
|
13
|
-
[project.scripts]
|
|
14
|
-
offlinedemo = "offlinedemo:main"
|
|
15
|
-
|
|
16
|
-
[build-system]
|
|
17
|
-
requires = ["uv_build>=0.12.19,<0.13.0"]
|
|
18
|
-
build-backend = "uv_build"
|
|
@@ -1,17 +0,0 @@
|
|
|
1
|
-
[project]
|
|
2
|
-
name = "offlinedemo"
|
|
3
|
-
version = "0.2.0"
|
|
4
|
-
description = "Easy demonstration of models by offlineisbetter."
|
|
5
|
-
readme = "README.md"
|
|
6
|
-
authors = [
|
|
7
|
-
{ name = "offlineisbetter", email = "dev@offlineisbetter.com" }
|
|
8
|
-
]
|
|
9
|
-
requires-python = ">=3.13"
|
|
10
|
-
dependencies = []
|
|
11
|
-
|
|
12
|
-
[project.scripts]
|
|
13
|
-
offlinedemo = "offlinedemo:main"
|
|
14
|
-
|
|
15
|
-
[build-system]
|
|
16
|
-
requires = ["uv_build>=0.12.19,<0.13.0"]
|
|
17
|
-
build-backend = "uv_build"
|
|
File without changes
|