├── images
    ├── screen.png
    └── CLIP-search.png
├── .env.example
├── pyproject.toml
├── src
    ├── clip.py
    ├── clusterer.py
    ├── main.py
    └── db.py
├── README.md
├── .gitignore
└── requirements.txt


/images/screen.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/AkiRusProd/CLIP-search/HEAD/images/screen.png


--------------------------------------------------------------------------------
/images/CLIP-search.png:
--------------------------------------------------------------------------------
https://raw.githubusercontent.com/AkiRusProd/CLIP-search/HEAD/images/CLIP-search.png


--------------------------------------------------------------------------------
/.env.example:
--------------------------------------------------------------------------------
1 | DEFAULT_IMAGES_PATH = "Directory with images"
2 | HUGGINGFACE_HUB_CACHE = "Path to Huggingface cache"
3 | INDEX_PATH = "./embed_data"


--------------------------------------------------------------------------------
/pyproject.toml:
--------------------------------------------------------------------------------
 1 | [tool.poetry]
 2 | name = "clip-search"
 3 | version = "0.1.0"
 4 | description = ""
 5 | authors = ["AkiRusProd <akirusprod@gmail.com>"]
 6 | readme = "README.md"
 7 | 
 8 | [tool.poetry.dependencies]
 9 | python = "^3.10"
10 | faiss-cpu = "1.7.4"
11 | gradio = "4.24.0"
12 | numpy = "1.24.0"
13 | pandas = "1.3.5"
14 | Pillow = "9.5.0"
15 | torch = "2.0.0"
16 | tqdm = "4.65.0"
17 | transformers = "4.30.2"
18 | python-dotenv = "^1.0.1"
19 | 
20 | 
21 | 
22 | [build-system]
23 | requires = ["poetry-core"]
24 | build-backend = "poetry.core.masonry.api"
25 | 


--------------------------------------------------------------------------------
/src/clip.py:
--------------------------------------------------------------------------------
 1 | 
 2 | import torch
 3 | from transformers import CLIPProcessor, CLIPModel, CLIPTokenizer
 4 | 
 5 | 
 6 | class CLIPSearcher:
 7 |     def __init__(self, model_id = "openai/clip-vit-base-patch32", device = None):
 8 |         if device is None:
 9 |             self.device = "cuda" if torch.cuda.is_available() else "cpu"
10 |         else:
11 |             self.device = device
12 | 
13 |         self.model: CLIPModel = CLIPModel.from_pretrained(model_id).to(device)
14 |         self.tokenizer: CLIPTokenizer = CLIPTokenizer.from_pretrained(model_id)
15 |         self.processor: CLIPProcessor = CLIPProcessor.from_pretrained(model_id)
16 | 
17 |     def get_text_features(self, text):
18 |         inputs = self.tokenizer(text, return_tensors = "pt").to(self.device)
19 |         return self.model.get_text_features(**inputs).cpu().detach().numpy()
20 | 
21 |     def get_image_features(self, image):
22 |         inputs = self.processor(images=image, return_tensors="pt").to(self.device)
23 |         return self.model.get_image_features(**inputs).cpu().detach().numpy()
24 | 
25 | 
26 | 
27 | 
28 | 


--------------------------------------------------------------------------------
/src/clusterer.py:
--------------------------------------------------------------------------------
 1 | from pathlib import Path
 2 | import faiss
 3 | import numpy as np
 4 | 
 5 | class ImageIndexer:
 6 |     def __init__(self, index_path):
 7 |         self.index = None
 8 |         self.index_path = Path(index_path)
 9 | 
10 |     def fit(self, image_embeds, num_clusters=None, m=32):
11 |         if isinstance(image_embeds, list):
12 |             image_embeds = np.array(image_embeds)
13 |         if num_clusters is None:
14 |             num_clusters = min(round(np.sqrt(image_embeds.shape[0])), image_embeds.shape[0] // 39) # Emperically determined
15 | 
16 |         n_dims = image_embeds.shape[1]
17 | 
18 |         quantizer = faiss.IndexHNSWFlat(n_dims, m)
19 |         self.index = faiss.IndexIVFFlat(quantizer, n_dims, num_clusters, faiss.METRIC_INNER_PRODUCT) # METRIC_INNER_PRODUCT: the higher this value, the better; METRIC_L2: other way round
20 |         self.index.train(image_embeds)
21 |         self.index.add(image_embeds)
22 | 
23 |         faiss.write_index(self.index, str(self.index_path / 'faiss_clusters.index'))
24 | 
25 |         return self.index
26 | 
27 |     def predict(self, image_embeds, embed, k):
28 |         if isinstance(image_embeds, list):
29 |             image_embeds = np.array(image_embeds)
30 |         if self.index is None:
31 |             self.index = faiss.read_index(str(self.index_path / 'faiss_clusters.index'))
32 | 
33 |         D, I = self.index.search(embed, k)
34 |         return D[0], I[0], image_embeds[I[0]]
35 | 


--------------------------------------------------------------------------------
/README.md:
--------------------------------------------------------------------------------
 1 | # CLIP-search
 2 | This is an impelementation of image search engine using CLIP (Contrastive Language-Image Pre-Training)
 3 | 
 4 | 
 5 | CLIP is a model that links text and images in a common space and makes it possible to understand the semantic relationship between them. CLIP is trained on a large dataset and is trained to understand how text and images relate to each other. CLIP can be used to classify images, search for similar images, generate text descriptions for images, and other tasks related to text and images.
 6 | 
 7 | 
 8 | <p align="center">
 9 | <img src="images/CLIP-search.png" width=100%>
10 | </p>
11 | 
12 | The current implementation allows you to find images in a local directory based on either text or another image. The CLIP model first computes embeddings for the provided text or image. Then it compares these embeddings with the embeddings of images in the local directory to find similar images. Finally, it returns the top k similar images from the directory.
13 | 
14 | This implementation leverages the power of CLIP's embeddings to enable searching for visually or semantically similar images in a local directory. By computing embeddings for both the query and the images in the directory, it can efficiently compare and identify the most similar images based on their embeddings.
15 | 
16 | 
17 | The implementation provides two methods for finding similar images:
18 | 
19 | 1. Slow method: Computes the cosine similarity between the query and all images in the directory individually. Сan be computationally expensive and time-consuming for large directories.
20 | 
21 | 2. Fast method: Utilizes clustering to find similar images by comparing the query with cluster centroids. Can be less accurate but faster. Speed matters for large directories.
22 | 
23 | You can choose between these methods based on your requirements for speed and accuracy.
24 | 
25 | 
26 | 
27 | 
28 | This implementation utilizes the [Hugging Face CLIP model](https://huggingface.co/openai/clip-vit-base-patch32). The [Faiss](https://github.com/facebookresearch/faiss) library is used for clustering, and the web UI is built using [Gradio](https://gradio.app/).
29 | 
30 | <p align="center">
31 | <img src="images/screen.png" width=100%>
32 | </p>
33 | 
34 | 
35 | Read more about CLIP at https://github.com/openai/CLIP.
36 | 


--------------------------------------------------------------------------------
/src/main.py:
--------------------------------------------------------------------------------
 1 | import os
 2 | import gradio as gr
 3 | from pathlib import Path
 4 | from PIL import Image
 5 | from db import SearchMechanism
 6 | from clip import CLIPSearcher
 7 | from clusterer import ImageIndexer
 8 | from dotenv import dotenv_values
 9 | 
10 | env = dotenv_values('.env')
11 | 
12 | path = env['DEFAULT_IMAGES_PATH']
13 | index_path = env['INDEX_PATH']
14 | 
15 | # your models cache will be stored here
16 | os.environ['HUGGINGFACE_HUB_CACHE'] = env['HUGGINGFACE_HUB_CACHE']
17 | 
18 | 
19 | def search_by_text(text, top_k, use_cluster_search):
20 |     top_k_df = search_mechanism.query_by_text(text, top_k, use_cluster_search)
21 |     images = [Image.open(row['image_path']) for _, row in top_k_df.iterrows()]
22 | 
23 |     return images
24 | 
25 | def search_by_image(image, top_k, use_cluster_search):
26 |     top_k_df = search_mechanism.query_by_image(image, top_k, use_cluster_search)
27 |     images = [Image.open(row['image_path']) for _, row in top_k_df.iterrows()]
28 | 
29 |     return images
30 | 
31 | def scan_dir(path):
32 |     if path is None or not os.path.exists(path):
33 |         return gr.Info("Path does not exist")
34 | 
35 |     search_mechanism.scan_directory(Path(path))
36 | 
37 |     return path
38 | 
39 | clip_searcher = CLIPSearcher()
40 | image_indexer = ImageIndexer(index_path)
41 | search_mechanism = SearchMechanism(clip_searcher, image_indexer)
42 | 
43 | with gr.Blocks() as webui:
44 |     gr.Markdown("CLIP Searcher")
45 | 
46 |     with gr.Row():
47 |         with gr.Column():
48 |             text = gr.Textbox(label = "Text", info = "Text to search")
49 | 
50 |             with gr.Row():
51 |                 image = gr.Image(label = "Image")
52 | 
53 |                 with gr.Column():
54 |                     top_k_slider = gr.Slider(label="Top K", minimum=1, maximum=50, step=1, value=5, info = "Top K closest results to the query")
55 |                     use_cluster_search = gr.Checkbox(label="Use cluster search", value=False, info="Faster embedding search using clusters, may be less accurate")
56 |                     search_by_text_btn = gr.Button("Search by text")
57 |                     search_by_image_btn = gr.Button("Search by image")
58 | 
59 |             path = gr.Textbox(label = "Path", info = "Path with images to scan", value=path)
60 |             scan_dir_btn = gr.Button("Scan directory", variant="primary")
61 | 
62 |   
63 |         gallery = gr.Gallery(label = "Gallery", show_label=False, columns=3, rows = 2, height="auto", preview = False)
64 | 
65 |     search_by_text_btn.click(
66 |         search_by_text, 
67 |         inputs = [text, top_k_slider, use_cluster_search], 
68 |         outputs = gallery
69 |     )
70 | 
71 |     search_by_image_btn.click(
72 |         search_by_image,
73 |         inputs = [image, top_k_slider, use_cluster_search],
74 |         outputs = gallery
75 |     )
76 | 
77 |     scan_dir_btn.click(
78 |         scan_dir,
79 |         inputs = [path],
80 |         outputs = path
81 |     )
82 | 
83 | webui.queue()
84 | webui.launch()


--------------------------------------------------------------------------------
/src/db.py:
--------------------------------------------------------------------------------
 1 | from pathlib import Path
 2 | import pandas as pd
 3 | import numpy as np
 4 | import torch
 5 | from tqdm import tqdm
 6 | from clip import CLIPSearcher
 7 | from clusterer import ImageIndexer
 8 | from PIL import Image
 9 | 
10 | tqdm.pandas()
11 | 
12 | 
13 | class SearchMechanism:
14 |     def __init__(self, clip_searcher: CLIPSearcher, image_indexer: ImageIndexer) -> None:
15 |         self.clip_searcher = clip_searcher
16 |         self.image_indexer = image_indexer
17 |         self.index_path = Path(self.image_indexer.index_path)
18 | 
19 |         self.df: pd.DataFrame = None
20 |         self.df_image_embeds: list = None
21 | 
22 |         self.load_db()
23 | 
24 |     def load_db(self):
25 |         try:
26 |             if not self.index_path.exists():
27 |                 self.index_path.mkdir(parents=True, exist_ok=True)
28 | 
29 |             df_path = self.index_path / 'df.csv'
30 |             embeds_path = self.index_path / 'df_image_embeds.npy'
31 | 
32 |             if df_path.exists() and embeds_path.exists():
33 |                 self.df = pd.read_csv(df_path, sep='\t')
34 |                 self.df_image_embeds = [x.flatten() for x in np.load(embeds_path)]
35 |         except Exception as e:
36 |             print(f"Error loading database: {e}")
37 | 
38 |     def scan_directory(self, path: Path):
39 |         if path is None or not path.exists():
40 |             raise Exception("Path does not exist")
41 | 
42 |         df = pd.DataFrame(columns=['image_path'])
43 |         df_image_embeds = []
44 | 
45 |         image_files = list(path.iterdir())
46 |         for i, img_path in tqdm(enumerate(image_files), total=len(image_files)):
47 |             df.loc[i, 'image_path'] = str(img_path)
48 |             df_image_embeds.append(self.clip_searcher.get_image_features(Image.open(img_path)).flatten())
49 | 
50 |         df.to_csv(self.index_path / 'df.csv', sep='\t', index=False)
51 |         np.save(self.index_path / 'df_image_embeds.npy', df_image_embeds)
52 | 
53 |         self.image_indexer.fit(df_image_embeds)
54 | 
55 |         self.load_db()
56 | 
57 |     def query_by_embeds(self, embeds: np.ndarray, top_k: int = 5, use_cluster_search: bool = False):
58 |         if self.df is None or self.df_image_embeds is None:
59 |             return
60 | 
61 |         df = self.df
62 | 
63 |         if not use_cluster_search:
64 |             df['cos_sim'] = pd.Series(self.df_image_embeds).progress_apply(
65 |                 lambda x: torch.nn.functional.cosine_similarity(torch.tensor(x), torch.tensor(embeds))
66 |             )
67 |             df = df.sort_values(by='cos_sim', ascending=False).head(top_k)
68 | 
69 |             return df.reset_index(drop=True)
70 |         else:
71 |             score, ids, _ = self.image_indexer.predict(self.df_image_embeds, embeds, top_k)
72 | 
73 |             ids = [id for id in ids if id != -1]
74 |             score = score[:len(ids)]
75 | 
76 |             df = df.iloc[ids].copy()
77 |             df['score'] = score
78 | 
79 |             df = df.sort_values(by='score', ascending=False)
80 |             
81 |             return df.reset_index(drop=True)
82 | 
83 |     def query_by_text(self, text: str, top_k: int = 5, use_cluster_search: bool = False):
84 |         text_embeds = self.clip_searcher.get_text_features(text)
85 |         
86 |         return self.query_by_embeds(text_embeds, top_k, use_cluster_search)
87 | 
88 |     def query_by_image(self, image, top_k: int = 5, use_cluster_search: bool = False):
89 |         if isinstance(image, str):
90 |             image = Image.open(image)
91 | 
92 |         image_embeds = self.clip_searcher.get_image_features(image)
93 |         
94 |         return self.query_by_embeds(image_embeds, top_k, use_cluster_search)
95 | 


--------------------------------------------------------------------------------
/.gitignore:
--------------------------------------------------------------------------------
  1 | # Byte-compiled / optimized / DLL files
  2 | __pycache__/
  3 | *.py[cod]
  4 | *$py.class
  5 | 
  6 | # C extensions
  7 | *.so
  8 | 
  9 | # Distribution / packaging
 10 | .Python
 11 | build/
 12 | develop-eggs/
 13 | dist/
 14 | downloads/
 15 | eggs/
 16 | .eggs/
 17 | lib/
 18 | lib64/
 19 | parts/
 20 | sdist/
 21 | var/
 22 | wheels/
 23 | share/python-wheels/
 24 | *.egg-info/
 25 | .installed.cfg
 26 | *.egg
 27 | MANIFEST
 28 | 
 29 | # PyInstaller
 30 | #  Usually these files are written by a python script from a template
 31 | #  before PyInstaller builds the exe, so as to inject date/other infos into it.
 32 | *.manifest
 33 | *.spec
 34 | 
 35 | # Installer logs
 36 | pip-log.txt
 37 | pip-delete-this-directory.txt
 38 | 
 39 | # Unit test / coverage reports
 40 | htmlcov/
 41 | .tox/
 42 | .nox/
 43 | .coverage
 44 | .coverage.*
 45 | .cache
 46 | nosetests.xml
 47 | coverage.xml
 48 | *.cover
 49 | *.py,cover
 50 | .hypothesis/
 51 | .pytest_cache/
 52 | cover/
 53 | 
 54 | # Translations
 55 | *.mo
 56 | *.pot
 57 | 
 58 | # Django stuff:
 59 | *.log
 60 | local_settings.py
 61 | db.sqlite3
 62 | db.sqlite3-journal
 63 | 
 64 | # Flask stuff:
 65 | instance/
 66 | .webassets-cache
 67 | 
 68 | # Scrapy stuff:
 69 | .scrapy
 70 | 
 71 | # Sphinx documentation
 72 | docs/_build/
 73 | 
 74 | # PyBuilder
 75 | .pybuilder/
 76 | target/
 77 | 
 78 | # Jupyter Notebook
 79 | .ipynb_checkpoints
 80 | 
 81 | # IPython
 82 | profile_default/
 83 | ipython_config.py
 84 | 
 85 | # pyenv
 86 | #   For a library or package, you might want to ignore these files since the code is
 87 | #   intended to run in multiple environments; otherwise, check them in:
 88 | # .python-version
 89 | 
 90 | # pipenv
 91 | #   According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control.
 92 | #   However, in case of collaboration, if having platform-specific dependencies or dependencies
 93 | #   having no cross-platform support, pipenv may install dependencies that don't work, or not
 94 | #   install all needed dependencies.
 95 | #Pipfile.lock
 96 | 
 97 | # poetry
 98 | #   Similar to Pipfile.lock, it is generally recommended to include poetry.lock in version control.
 99 | #   This is especially recommended for binary packages to ensure reproducibility, and is more
100 | #   commonly ignored for libraries.
101 | #   https://python-poetry.org/docs/basic-usage/#commit-your-poetrylock-file-to-version-control
102 | #poetry.lock
103 | 
104 | # pdm
105 | #   Similar to Pipfile.lock, it is generally recommended to include pdm.lock in version control.
106 | #pdm.lock
107 | #   pdm stores project-wide configurations in .pdm.toml, but it is recommended to not include it
108 | #   in version control.
109 | #   https://pdm.fming.dev/#use-with-ide
110 | .pdm.toml
111 | 
112 | # PEP 582; used by e.g. github.com/David-OConnor/pyflow and github.com/pdm-project/pdm
113 | __pypackages__/
114 | 
115 | # Celery stuff
116 | celerybeat-schedule
117 | celerybeat.pid
118 | 
119 | # SageMath parsed files
120 | *.sage.py
121 | 
122 | # Environments
123 | .env
124 | .venv
125 | env/
126 | venv/
127 | ENV/
128 | env.bak/
129 | venv.bak/
130 | 
131 | # Spyder project settings
132 | .spyderproject
133 | .spyproject
134 | 
135 | # Rope project settings
136 | .ropeproject
137 | 
138 | # mkdocs documentation
139 | /site
140 | 
141 | # mypy
142 | .mypy_cache/
143 | .dmypy.json
144 | dmypy.json
145 | 
146 | # Pyre type checker
147 | .pyre/
148 | 
149 | # pytype static type analyzer
150 | .pytype/
151 | 
152 | # Cython debug symbols
153 | cython_debug/
154 | 
155 | # PyCharm
156 | #  JetBrains specific template is maintained in a separate JetBrains.gitignore that can
157 | #  be found at https://github.com/github/gitignore/blob/main/Global/JetBrains.gitignore
158 | #  and can be added to the global gitignore or merged into this file.  For a more nuclear
159 | #  option (not recommended) you can uncomment the following to ignore the entire idea folder.
160 | #.idea/
161 | 
162 | test.py
163 | embed_data/


--------------------------------------------------------------------------------
/requirements.txt:
--------------------------------------------------------------------------------
 1 | aiofiles==23.2.1 ; python_version >= "3.10" and python_version < "4.0"
 2 | altair==5.3.0 ; python_version >= "3.10" and python_version < "4.0"
 3 | annotated-types==0.6.0 ; python_version >= "3.10" and python_version < "4.0"
 4 | anyio==4.3.0 ; python_version >= "3.10" and python_version < "4.0"
 5 | attrs==23.2.0 ; python_version >= "3.10" and python_version < "4.0"
 6 | certifi==2024.2.2 ; python_version >= "3.10" and python_version < "4.0"
 7 | charset-normalizer==3.3.2 ; python_version >= "3.10" and python_version < "4.0"
 8 | click==8.1.7 ; python_version >= "3.10" and python_version < "4.0" and sys_platform != "emscripten"
 9 | cmake==3.29.0.1 ; platform_system == "Linux" and platform_machine == "x86_64" and python_version >= "3.10" and python_version < "4.0"
10 | colorama==0.4.6 ; python_version >= "3.10" and python_version < "4.0" and platform_system == "Windows"
11 | contourpy==1.2.0 ; python_version >= "3.10" and python_version < "4.0"
12 | cycler==0.12.1 ; python_version >= "3.10" and python_version < "4.0"
13 | exceptiongroup==1.2.0 ; python_version >= "3.10" and python_version < "3.11"
14 | faiss-cpu==1.7.4 ; python_version >= "3.10" and python_version < "4.0"
15 | fastapi==0.110.0 ; python_version >= "3.10" and python_version < "4.0"
16 | ffmpy==0.3.2 ; python_version >= "3.10" and python_version < "4.0"
17 | filelock==3.13.3 ; python_version >= "3.10" and python_version < "4.0"
18 | fonttools==4.50.0 ; python_version >= "3.10" and python_version < "4.0"
19 | fsspec==2024.3.1 ; python_version >= "3.10" and python_version < "4.0"
20 | gradio-client==0.14.0 ; python_version >= "3.10" and python_version < "4.0"
21 | gradio==4.24.0 ; python_version >= "3.10" and python_version < "4.0"
22 | h11==0.14.0 ; python_version >= "3.10" and python_version < "4.0"
23 | httpcore==1.0.5 ; python_version >= "3.10" and python_version < "4.0"
24 | httpx==0.27.0 ; python_version >= "3.10" and python_version < "4.0"
25 | huggingface-hub==0.22.2 ; python_version >= "3.10" and python_version < "4.0"
26 | idna==3.6 ; python_version >= "3.10" and python_version < "4.0"
27 | importlib-resources==6.4.0 ; python_version >= "3.10" and python_version < "4.0"
28 | jinja2==3.1.3 ; python_version >= "3.10" and python_version < "4.0"
29 | jsonschema-specifications==2023.12.1 ; python_version >= "3.10" and python_version < "4.0"
30 | jsonschema==4.21.1 ; python_version >= "3.10" and python_version < "4.0"
31 | kiwisolver==1.4.5 ; python_version >= "3.10" and python_version < "4.0"
32 | lit==18.1.2 ; platform_system == "Linux" and platform_machine == "x86_64" and python_version >= "3.10" and python_version < "4.0"
33 | markdown-it-py==3.0.0 ; python_version >= "3.10" and python_version < "4.0" and sys_platform != "emscripten"
34 | markupsafe==2.1.5 ; python_version >= "3.10" and python_version < "4.0"
35 | matplotlib==3.8.3 ; python_version >= "3.10" and python_version < "4.0"
36 | mdurl==0.1.2 ; python_version >= "3.10" and python_version < "4.0" and sys_platform != "emscripten"
37 | mpmath==1.3.0 ; python_version >= "3.10" and python_version < "4.0"
38 | networkx==3.2.1 ; python_version >= "3.10" and python_version < "4.0"
39 | numpy==1.24.0 ; python_version >= "3.10" and python_version < "4.0"
40 | nvidia-cublas-cu11==11.10.3.66 ; platform_system == "Linux" and platform_machine == "x86_64" and python_version >= "3.10" and python_version < "4.0"
41 | nvidia-cuda-cupti-cu11==11.7.101 ; platform_system == "Linux" and platform_machine == "x86_64" and python_version >= "3.10" and python_version < "4.0"
42 | nvidia-cuda-nvrtc-cu11==11.7.99 ; platform_system == "Linux" and platform_machine == "x86_64" and python_version >= "3.10" and python_version < "4.0"
43 | nvidia-cuda-runtime-cu11==11.7.99 ; platform_system == "Linux" and platform_machine == "x86_64" and python_version >= "3.10" and python_version < "4.0"
44 | nvidia-cudnn-cu11==8.5.0.96 ; platform_system == "Linux" and platform_machine == "x86_64" and python_version >= "3.10" and python_version < "4.0"
45 | nvidia-cufft-cu11==10.9.0.58 ; platform_system == "Linux" and platform_machine == "x86_64" and python_version >= "3.10" and python_version < "4.0"
46 | nvidia-curand-cu11==10.2.10.91 ; platform_system == "Linux" and platform_machine == "x86_64" and python_version >= "3.10" and python_version < "4.0"
47 | nvidia-cusolver-cu11==11.4.0.1 ; platform_system == "Linux" and platform_machine == "x86_64" and python_version >= "3.10" and python_version < "4.0"
48 | nvidia-cusparse-cu11==11.7.4.91 ; platform_system == "Linux" and platform_machine == "x86_64" and python_version >= "3.10" and python_version < "4.0"
49 | nvidia-nccl-cu11==2.14.3 ; platform_system == "Linux" and platform_machine == "x86_64" and python_version >= "3.10" and python_version < "4.0"
50 | nvidia-nvtx-cu11==11.7.91 ; platform_system == "Linux" and platform_machine == "x86_64" and python_version >= "3.10" and python_version < "4.0"
51 | orjson==3.10.0 ; python_version >= "3.10" and python_version < "4.0"
52 | packaging==24.0 ; python_version >= "3.10" and python_version < "4.0"
53 | pandas==1.3.5 ; python_version >= "3.10" and python_version < "4.0"
54 | pillow==9.5.0 ; python_version >= "3.10" and python_version < "4.0"
55 | pydantic-core==2.16.3 ; python_version >= "3.10" and python_version < "4.0"
56 | pydantic==2.6.4 ; python_version >= "3.10" and python_version < "4.0"
57 | pydub==0.25.1 ; python_version >= "3.10" and python_version < "4.0"
58 | pygments==2.17.2 ; python_version >= "3.10" and python_version < "4.0" and sys_platform != "emscripten"
59 | pyparsing==3.1.2 ; python_version >= "3.10" and python_version < "4.0"
60 | python-dateutil==2.9.0.post0 ; python_version >= "3.10" and python_version < "4.0"
61 | python-dotenv==1.0.1 ; python_version >= "3.10" and python_version < "4.0"
62 | python-multipart==0.0.9 ; python_version >= "3.10" and python_version < "4.0"
63 | pytz==2024.1 ; python_version >= "3.10" and python_version < "4.0"
64 | pyyaml==6.0.1 ; python_version >= "3.10" and python_version < "4.0"
65 | referencing==0.34.0 ; python_version >= "3.10" and python_version < "4.0"
66 | regex==2023.12.25 ; python_version >= "3.10" and python_version < "4.0"
67 | requests==2.31.0 ; python_version >= "3.10" and python_version < "4.0"
68 | rich==13.7.1 ; python_version >= "3.10" and python_version < "4.0" and sys_platform != "emscripten"
69 | rpds-py==0.18.0 ; python_version >= "3.10" and python_version < "4.0"
70 | ruff==0.3.4 ; python_version >= "3.10" and python_version < "4.0" and sys_platform != "emscripten"
71 | safetensors==0.4.2 ; python_version >= "3.10" and python_version < "4.0"
72 | semantic-version==2.10.0 ; python_version >= "3.10" and python_version < "4.0"
73 | setuptools==69.2.0 ; platform_system == "Linux" and platform_machine == "x86_64" and python_version >= "3.10" and python_version < "4.0"
74 | shellingham==1.5.4 ; python_version >= "3.10" and python_version < "4.0" and sys_platform != "emscripten"
75 | six==1.16.0 ; python_version >= "3.10" and python_version < "4.0"
76 | sniffio==1.3.1 ; python_version >= "3.10" and python_version < "4.0"
77 | starlette==0.36.3 ; python_version >= "3.10" and python_version < "4.0"
78 | sympy==1.12 ; python_version >= "3.10" and python_version < "4.0"
79 | tokenizers==0.13.3 ; python_version >= "3.10" and python_version < "4.0"
80 | tomlkit==0.12.0 ; python_version >= "3.10" and python_version < "4.0"
81 | toolz==0.12.1 ; python_version >= "3.10" and python_version < "4.0"
82 | torch==2.0.0 ; python_version >= "3.10" and python_version < "4.0"
83 | tqdm==4.65.0 ; python_version >= "3.10" and python_version < "4.0"
84 | transformers==4.30.2 ; python_version >= "3.10" and python_version < "4.0"
85 | triton==2.0.0 ; platform_system == "Linux" and platform_machine == "x86_64" and python_version >= "3.10" and python_version < "4.0"
86 | typer-cli==0.12.0 ; python_version >= "3.10" and python_version < "4.0" and sys_platform != "emscripten"
87 | typer-slim[standard]==0.12.0 ; python_version >= "3.10" and python_version < "4.0" and sys_platform != "emscripten"
88 | typer[all]==0.12.0 ; python_version >= "3.10" and python_version < "4.0" and sys_platform != "emscripten"
89 | typing-extensions==4.10.0 ; python_version >= "3.10" and python_version < "4.0"
90 | urllib3==2.2.1 ; python_version >= "3.10" and python_version < "4.0"
91 | uvicorn==0.29.0 ; python_version >= "3.10" and python_version < "4.0" and sys_platform != "emscripten"
92 | websockets==11.0.3 ; python_version >= "3.10" and python_version < "4.0"
93 | wheel==0.43.0 ; platform_system == "Linux" and platform_machine == "x86_64" and python_version >= "3.10" and python_version < "4.0"
94 | 


--------------------------------------------------------------------------------