testings for gallery 2.0

Archive/agraphTest.py

@@ -0,0 +1,170 @@
+import os
+
+import streamlit as st
+import torch
+import pandas as pd
+import numpy as np
+
+from datasets import load_dataset, Dataset, load_from_disk
+from huggingface_hub import login
+from streamlit_agraph import agraph, Node, Edge, Config
+from sklearn.manifold import TSNE
+
+
+@st.cache_data
+def load_hf_dataset():
+    # login to huggingface
+    login(token=os.environ.get("HF_TOKEN"))
+
+    # load from huggingface
+    roster = pd.DataFrame(load_dataset('MAPS-research/GEMRec-Roster', split='train'))
+    promptBook = pd.DataFrame(load_dataset('MAPS-research/GEMRec-Metadata', split='train'))
+
+    # process dataset
+    roster = roster[['model_id', 'model_name', 'modelVersion_id', 'modelVersion_name',
+                                                       'model_download_count']].drop_duplicates().reset_index(drop=True)
+
+    # add 'custom_score_weights' column to promptBook if not exist
+    if 'weighted_score_sum' not in promptBook.columns:
+        promptBook.loc[:, 'weighted_score_sum'] = 0
+
+    # merge roster and promptbook
+    promptBook = promptBook.merge(roster[['model_id', 'model_name', 'modelVersion_id', 'modelVersion_name', 'model_download_count']],
+                                                                    on=['model_id', 'modelVersion_id'], how='left')
+
+    # add column to record current row index
+    promptBook.loc[:, 'row_idx'] = promptBook.index
+
+    return roster, promptBook
+
+
+@st.cache_data
+def calc_tsne(prompt_id):
+    print('==> loading feats')
+    feats = {}
+    for pt in os.listdir('../data/feats'):
+        if pt.split('.')[-1] == 'pt' and pt.split('.')[0].isdigit():
+            feats[pt.split('.')[0]] = torch.load(os.path.join('../data/feats', pt))
+
+    print('==> applying t-SNE')
+    # apply t-SNE to entries in each feat in feats to get 2D coordinates
+    tsne = TSNE(n_components=2, random_state=0)
+    # for k, v in tqdm(feats.items()):
+    #     feats[k]['tsne'] = tsne.fit_transform(v['all'].numpy())
+    # prompt_id = '90'
+    feats[prompt_id]['tsne'] = tsne.fit_transform(feats[prompt_id]['all'].numpy())
+
+    feats_df = pd.DataFrame(feats[prompt_id]['tsne'], columns=['x', 'y'])
+    feats_df['prompt_id'] = prompt_id
+
+    keys = []
+    for k in feats[prompt_id].keys():
+        if k != 'all' and k != 'tsne':
+            keys.append(int(k.item()))
+
+    feats_df['modelVersion_id'] = keys
+
+
+    return feats_df
+
+    # print(feats[prompt_id]['tsne'])
+
+
+if __name__ == '__main__':
+    st.set_page_config(layout="wide")
+
+    # load dataset
+    roster, promptBook = load_hf_dataset()
+    # prompt_id = '20'
+
+    with st.sidebar:
+        st.write('## Select Prompt')
+        prompts = promptBook['prompt_id'].unique().tolist()
+        # sort prompts by prompt_id
+        prompts.sort()
+        prompt_id = st.selectbox('Select Prompt', prompts, index=0)
+        physics = st.checkbox('Enable Physics')
+
+    feats_df = calc_tsne(str(prompt_id))
+
+    # keys = []
+    # for k in feats[prompt_id].keys():
+    #     if k != 'all' and k != 'tsne':
+    #         keys.append(int(k.item()))
+
+    # print(keys)
+
+    data = []
+    for idx in feats_df.index:
+        modelVersion_id = feats_df.loc[idx, 'modelVersion_id']
+        image_id = promptBook[(promptBook['modelVersion_id'] == modelVersion_id) & (
+                    promptBook['prompt_id'] == int(prompt_id))].reset_index(drop=True).loc[0, 'image_id']
+        image_url = f"https://modelcofferbucket.s3-accelerate.amazonaws.com/{image_id}.png"
+        scale = 50
+        data.append((feats_df.loc[idx, 'x'] * scale, feats_df.loc[idx, 'y'] * scale, image_url))
+
+        image_size = promptBook[(promptBook['image_id'] == image_id)].reset_index(drop=True).loc[0, 'size'].split('x')
+
+    nodes = []
+    edges = []
+
+    for d in data:
+        nodes.append( Node(id=d[2],
+                           # label=str(items.loc[idx, 'model_name']),
+                           size=20,
+                           shape="image",
+                           image=d[2],
+                           x=[d[0]],
+                           y=[d[1]],
+                           fixed=False if physics else True,
+                           color={'background': '#00000', 'border': '#ffffff'},
+                           shadow={'enabled': True, 'color': 'rgba(0,0,0,0.4)', 'size': 10, 'x': 1, 'y': 1},
+                           # borderWidth=1,
+                           # shapeProperties={'useBorderWithImage': True},
+                           )
+                    )
+
+
+    # nodes.append( Node(id="Spiderman",
+    #                    label="Peter Parker",
+    #                    size=25,
+    #                    shape="circularImage",
+    #                    image="http://marvel-force-chart.surge.sh/marvel_force_chart_img/top_spiderman.png")
+    #             ) # includes **kwargs
+    # nodes.append( Node(id="Captain_Marvel",
+    #                    label="Carol Danvers",
+    #                    fixed=True,
+    #                    size=25,
+    #                    shape="circularImage",
+    #                    image="http://marvel-force-chart.surge.sh/marvel_force_chart_img/top_captainmarvel.png")
+    #             )
+    # edges.append( Edge(source="Captain_Marvel",
+    #                    label="friend_of",
+    #                    target="Spiderman",
+    #                    length=200,
+    #                    # **kwargs
+    #                    )
+    #             )
+    #
+    config = Config(width='100%',
+                    height=800,
+                    directed=True,
+                    physics=physics,
+                    hierarchical=False,
+                    # **kwargs
+                    )
+
+    cols = st.columns([3, 1], gap='large')
+
+    with cols[0]:
+            return_value = agraph(nodes=nodes,
+                                  edges=edges,
+                                  config=config)
+
+    # st.write(return_value)
+
+    with cols[1]:
+        try:
+            st.image(return_value, use_column_width=True)
+        except:
+            st.write('No image selected')

Archive/bokehTest.py

@@ -0,0 +1,182 @@
+import os
+
+import streamlit as st
+import torch
+import pandas as pd
+import numpy as np
+import requests
+
+from bokeh.plotting import figure, show
+from bokeh.models import HoverTool, ColumnDataSource, CustomJSHover
+from bokeh.embed import file_html
+from bokeh.resources import CDN  # Import CDN here
+from datasets import load_dataset, Dataset, load_from_disk
+from huggingface_hub import login
+from sklearn.manifold import TSNE
+from tqdm import tqdm
+
+
+@st.cache_data
+def load_hf_dataset():
+    # login to huggingface
+    login(token=os.environ.get("HF_TOKEN"))
+
+    # load from huggingface
+    roster = pd.DataFrame(load_dataset('MAPS-research/GEMRec-Roster', split='train'))
+    promptBook = pd.DataFrame(load_dataset('MAPS-research/GEMRec-Metadata', split='train'))
+
+    # process dataset
+    roster = roster[['model_id', 'model_name', 'modelVersion_id', 'modelVersion_name',
+                                                       'model_download_count']].drop_duplicates().reset_index(drop=True)
+
+    # add 'custom_score_weights' column to promptBook if not exist
+    if 'weighted_score_sum' not in promptBook.columns:
+        promptBook.loc[:, 'weighted_score_sum'] = 0
+
+    # merge roster and promptbook
+    promptBook = promptBook.merge(roster[['model_id', 'model_name', 'modelVersion_id', 'modelVersion_name', 'model_download_count']],
+                                                                    on=['model_id', 'modelVersion_id'], how='left')
+
+    # add column to record current row index
+    promptBook.loc[:, 'row_idx'] = promptBook.index
+
+    return roster, promptBook
+
+def show_with_bokeh(data, streamlit=False):
+    # Extract x, y coordinates and image URLs
+    x_coords, y_coords, image_urls = zip(*data)
+
+    # Create a ColumnDataSource
+    source = ColumnDataSource(data=dict(x=x_coords, y=y_coords, image=image_urls))
+
+    # Create a figure
+    p = figure(width=800, height=600)
+
+    # Add scatter plot
+    scatter = p.scatter(x='x', y='y', size=20, source=source)
+
+    # Define hover tool
+    hover = HoverTool()
+    # hover.tooltips = """
+    #         <div>
+    #             <iframe src="@image" width="512" height="512"></iframe>
+    #         </div>
+    #     """
+    # hover.formatters = {'@image': CustomJSHover(code="""
+    #         const index = cb_data.index;
+    #         const url = cb_data.source.data['image'][index];
+    #         return '<iframe src="' + url + '" width="512" height="512"></iframe>';
+    #     """)}
+
+    hover.tooltips = """
+            <div>
+                <img src="@image" style='object-fit: contain'; height=100%">
+            </div>
+        """
+    hover.formatters = {'@image': CustomJSHover(code="""
+            const index = cb_data.index;
+            const url = cb_data.source.data['image'][index];
+            return '<img src="' + url + '">';
+        """)}
+
+    p.add_tools(hover)
+
+    # Generate HTML with the plot
+    html = file_html(p, CDN, "Interactive Scatter Plot with Hover Images")
+
+    # Save the HTML file or show it
+    # with open("scatter_plot_with_hover_images.html", "w") as f:
+    #     f.write(html)
+
+    if streamlit:
+        st.bokeh_chart(p, use_container_width=True)
+    else:
+        show(p)
+
+
+def show_with_bokeh_2(data, image_size=[40, 40], streamlit=False):
+    # Extract x, y coordinates and image URLs
+    x_coords, y_coords, image_urls = zip(*data)
+
+    # Create a ColumnDataSource
+    source = ColumnDataSource(data=dict(x=x_coords, y=y_coords, image=image_urls))
+
+    # Create a figure
+    p = figure(width=800, height=600, aspect_ratio=1.0)
+
+    # Add image glyphs
+    # image_size = 40  # Adjust this size as needed
+    scale = 0.1
+    image_size = [int(image_size[0])*scale, int(image_size[1])*scale]
+    print(image_size)
+    p.image_url(url='image', x='x', y='y', source=source, w=image_size[0], h=image_size[1], anchor="center")
+
+    # Define hover tool
+    hover = HoverTool()
+    hover.tooltips = """
+            <div>
+                <img src="@image" style='object-fit: contain'; height=100%'">
+            </div>
+        """
+    p.add_tools(hover)
+
+    # Generate HTML with the plot
+    html = file_html(p, CDN, "Scatter Plot with Images")
+
+    # Save the HTML file or show it
+    # with open("scatter_plot_with_images.html", "w") as f:
+    #     f.write(html)
+
+    if streamlit:
+        st.bokeh_chart(p, use_container_width=True)
+    else:
+        show(p)
+
+
+if __name__ == '__main__':
+    # load dataset
+    roster, promptBook = load_hf_dataset()
+
+    print('==> loading feats')
+    feats = {}
+    for pt in os.listdir('../data/feats'):
+        if pt.split('.')[-1] == 'pt' and pt.split('.')[0].isdigit():
+            feats[pt.split('.')[0]] = torch.load(os.path.join('../data/feats', pt))
+
+    print('==> applying t-SNE')
+    # apply t-SNE to entries in each feat in feats to get 2D coordinates
+    tsne = TSNE(n_components=2, random_state=0)
+    # for k, v in tqdm(feats.items()):
+    #     feats[k]['tsne'] = tsne.fit_transform(v['all'].numpy())
+    prompt_id = '49'
+    feats[prompt_id]['tsne'] = tsne.fit_transform(feats[prompt_id]['all'].numpy())
+
+    print(feats[prompt_id]['tsne'])
+
+    keys = []
+    for k in feats[prompt_id].keys():
+        if k != 'all' and k != 'tsne':
+            keys.append(int(k.item()))
+
+    print(keys)
+
+    data = []
+    for idx in range(len(keys)):
+        modelVersion_id = keys[idx]
+        image_id = promptBook[(promptBook['modelVersion_id'] == modelVersion_id) & (promptBook['prompt_id'] == int(prompt_id))].reset_index(drop=True).loc[0, 'image_id']
+        image_url = f"https://modelcofferbucket.s3-accelerate.amazonaws.com/{image_id}.png"
+        scale = 50
+        data.append((feats[prompt_id]['tsne'][idx][0]*scale, feats[prompt_id]['tsne'][idx][1]*scale, image_url))
+
+        image_size = promptBook[(promptBook['image_id'] == image_id)].reset_index(drop=True).loc[0, 'size'].split('x')
+
+    # # Sample data: (x, y) coordinates and corresponding image URLs
+    # data = [
+    #     (2, 5, "https://www.crunchyroll.com/imgsrv/display/thumbnail/480x720/catalog/crunchyroll/669dae5dbea3d93bb5f1012078501976.jpeg"),
+    #     (4, 8, "https://i.pinimg.com/originals/40/6d/38/406d38957bc4fd12f34c5dfa3d73b86d.jpg"),
+    #     (7, 3, "https://i.pinimg.com/550x/76/27/d2/7627d227adc6fb5fb6662ebfb9d82d7e.jpg"),
+    #     # Add more data points and image URLs
+    # ]
+
+    # show_with_bokeh(data, streamlit=True)
+    show_with_bokeh_2(data, image_size=image_size, streamlit=True)

Archive/test.py

@@ -1,124 +0,0 @@
-import streamlit as st
-from streamlit_sortables import sort_items
-from torchvision import transforms
-from transformers import CLIPProcessor, CLIPModel
-from torchmetrics.multimodal import CLIPScore
-import torch
-import numpy as np
-import pandas as pd
-from tqdm import tqdm
-from datasets import load_dataset, Dataset, load_from_disk
-import os
-
-import clip
-
-def compute_clip_score(promptbook, device, drop_negative=False):
-    # if 'clip_score' in promptbook.columns:
-    #     print('==> Skipping CLIP-Score computation')
-    #     return
-    print('==> CLIP-Score computation started')
-    clip_scores = []
-    to_tensor = transforms.ToTensor()
-    # metric = CLIPScore(model_name_or_path='openai/clip-vit-base-patch16').to(DEVICE)
-    metric = CLIPScore(model_name_or_path='openai/clip-vit-large-patch14').to(device)
-    for i in tqdm(range(0, len(promptbook), BATCH_SIZE)):
-        images = []
-        prompts = list(promptbook.prompt.values[i:i+BATCH_SIZE])
-        for image in promptbook.image.values[i:i+BATCH_SIZE]:
-            images.append(to_tensor(image))
-        with torch.no_grad():
-            x = metric.processor(text=prompts, images=images, return_tensors='pt', padding=True)
-            img_features = metric.model.get_image_features(x['pixel_values'].to(device))
-            img_features = img_features / img_features.norm(p=2, dim=-1, keepdim=True)
-            txt_features = metric.model.get_text_features(x['input_ids'].to(device), x['attention_mask'].to(device))
-            txt_features = txt_features / txt_features.norm(p=2, dim=-1, keepdim=True)
-            scores = 100 * (img_features * txt_features).sum(axis=-1).detach().cpu()
-        if drop_negative:
-            scores = torch.max(scores, torch.zeros_like(scores))
-        clip_scores += [round(s.item(), 4) for s in scores]
-    promptbook['clip_score'] = np.asarray(clip_scores)
-    print('==> CLIP-Score computation completed')
-    return promptbook
-
-
-def compute_clip_score_hmd(promptbook):
-
-    metric_cpu = CLIPScore(model_name_or_path="openai/clip-vit-large-patch14").to('cpu')
-    metric_gpu = CLIPScore(model_name_or_path="openai/clip-vit-large-patch14").to('mps')
-
-    for idx in promptbook.index:
-        clip_score_hm = promptbook.loc[idx, 'clip_score']
-
-        with torch.no_grad():
-            image = promptbook.loc[idx, 'image']
-            image.save(f"./tmp/{promptbook.loc[idx, 'image_id']}.png")
-            image = transforms.ToTensor()(image)
-            image_cpu = torch.unsqueeze(image, dim=0).to('cpu')
-            image_gpu = torch.unsqueeze(image, dim=0).to('mps')
-
-            prompts = [promptbook.loc[idx, 'prompt']]
-            clip_score_cpu = metric_cpu(image_cpu, prompts)
-            clip_score_gpu = metric_gpu(image_gpu, prompts)
-
-        print(
-            f'==> clip_score_hm: {clip_score_hm:.4f}, clip_score_cpu: {clip_score_cpu:.4f}, clip_score_gpu: {clip_score_gpu:.4f}')
-
-def compute_clip_score_transformers(promptbook, device='cpu'):
-    model = CLIPModel.from_pretrained("openai/clip-vit-large-patch14")
-    processor = CLIPProcessor.from_pretrained("openai/clip-vit-large-patch14")
-
-    with torch.no_grad():
-        inputs = processor(text=promptbook.prompt.tolist(), images=promptbook.image.tolist(), return_tensors="pt", padding=True)
-        outputs = model(**inputs)
-        logits_per_image = outputs.logits_per_image
-
-    promptbook.loc[:, 'clip_score'] = logits_per_image[:, 0].tolist()
-    return promptbook
-
-def compute_clip_score_clip(promptbook, device='cpu'):
-    model, preprocess = clip.load("ViT-B/32", device=device)
-    with torch.no_grad():
-
-        for idx in promptbook.index:
-            # image_input = preprocess(promptbook.loc[idx, 'image']).unsqueeze(0).to(device)
-            image_inputs = preprocess(promptbook.image.tolist()).to(device)
-            text_inputs = torch.cat([clip.tokenize(promptbook.prompt.tolist()).to(device)]).to(device)
-
-            image_features = model.encode_image(image_inputs)
-            text_features = model.encode_text(text_inputs)
-
-
-            probs = logits_per_image.softmax(dim=-1).cpu().numpy()
-    promptbook.loc[:, 'clip_score'] = probs[:, 0].tolist()
-    return promptbook
-
-
-if __name__ == "__main__":
-    BATCH_SIZE = 200
-    # DEVICE = 'mps' if torch.has_mps else 'cpu'
-
-    print(torch.__version__)
-
-    images_ds = load_from_disk(os.path.join(os.pardir, 'data', 'promptbook'))
-    images_ds = images_ds.sort(['prompt_id', 'modelVersion_id'])
-    print(images_ds)
-    print(type(images_ds[0]['image']))
-    promptbook_hmd = pd.DataFrame(images_ds[:20])
-    promptbook_new = promptbook_hmd.drop(columns=['clip_score'])
-    promptbook_cpu = compute_clip_score(promptbook_new.copy(deep=True), device='cpu')
-    promptbook_mps = compute_clip_score(promptbook_new.copy(deep=True), device='mps')
-    promptbook_tra_cpu = compute_clip_score_transformers(promptbook_new.copy(deep=True))
-    promptbook_tra_mps = compute_clip_score_transformers(promptbook_new.copy(deep=True), device='mps')
-    #
-    for idx in promptbook_mps.index:
-        print(
-            'image id: ', promptbook_mps['image_id'][idx],
-            'mps: ', promptbook_mps['clip_score'][idx],
-            'cpu: ', promptbook_cpu['clip_score'][idx],
-            'tra cpu: ', promptbook_tra_cpu['clip_score'][idx],
-            'tra mps: ', promptbook_tra_mps['clip_score'][idx],
-            'hmd: ', promptbook_hmd['clip_score'][idx]
-        )
-    #
-    # compute_clip_score_hmd(promptbook_hmd)
-

Archive/test_form.py

@@ -1,39 +0,0 @@
-import streamlit as st
-
-
-def grid(col=3, row=4, name='grid1'):
-    cols = st.columns(col)
-    for i in range(row):
-        for j in range(col):
-            with cols[j]:
-                value = st.session_state.checked_dic[name].get(f"{name}_{i*col+j}", False)
-
-                check = st.checkbox(f"{i*col+j}", key=f"{name}_{i*col+j}", value=value)
-                if check:
-                    st.session_state.checked_dic[name][f"{name}_{i*col+j}"] = True
-                else:
-                    st.session_state.checked_dic[name][f"{name}_{i*col+j}"] = False
-
-
-def on_click():
-    for key in st.session_state:
-        if st.session_state[key] and key[-1].isdigit():
-            st.write(key)
-    # for key in st.session_state.checked_dic[name]:
-    #     if st.session_state.checked_dic[name][key]:
-    #         st.write(key)
-
-
-
-if __name__ == "__main__":
-    if 'checked_dic' not in st.session_state:
-        st.session_state.checked_dic = {'grid1': {}, 'grid2': {}}
-
-    name = st.selectbox('Select a grid', ['grid1', 'grid2'])
-
-    with st.form(f"{name}_form"):
-        grid(name=name)
-        submit_button = st.form_submit_button("Submit", on_click=on_click)
-
-
-

data/feats/1.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:211566fa05419c50f9713be092f343a4518a68a0ecdec783f957412d8924cda1
+size 639065

data/feats/10.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3ddfed333b3a19193d15c23f29d61187e5becc3aaf115c7ad14cbd860c69efcf
+size 639068

data/feats/11.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aa36101aba0c01e1e8a4ad47c506847d5540db36217ab31de59155f605713e70
+size 639068

data/feats/12.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6bc9518b972441e3e0d5144025112ad20e0a501914b68806d4ff1ef6e7e7b58a
+size 639068

data/feats/13.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:378aca12dff37e85f2204043bde53c015e85835bee288c317d809abf425b8789
+size 639068

data/feats/14.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:30b57b3a83a9ba6b78e161f266ab672e054e7772d763c43c1bac80aecd6c8541
+size 639068

data/feats/15.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d8b0115e79c8a4e6a6b9ea5c4985fb63d6241a571a64d752db64ac640f74ca34
+size 639068

data/feats/16.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e2a6bd80f3f5dd395a8b0d03e5b39e2e5c17d80f882289a51613a425d09a1ead
+size 639068

data/feats/17.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:79dab0b91e00271acb800da17e2d74eebb034cb8a95e14517e4716d5e45c6ad0
+size 639068

data/feats/18.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2fe951fd6f45722d23dc67f1680651978e04e0a2ac922bf32c089a70cfb87c42
+size 639068

data/feats/19.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ee7b8f1b35d51c87b76e315fd70f6eae65bd3f0722a09ccf8f9bedbe6c964ad8
+size 639068

data/feats/2.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:51df1dd4539225460d88d237881af030e0a8919ead4840ee12649902d2cb9e27
+size 639065

data/feats/20.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:18c90e42f00fc0fab451ad6abf1a1ebac69effeddb3996c53d9cb7f42c54ddac
+size 639068

data/feats/21.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e2291f89dbd9db4a440027b70f32007dfe7a30b71b7b724c8cd5227297b309e8
+size 639068

data/feats/22.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2a5b588af6935a96915b4aededf0aab030d5af2a4c90e82ac0f7f287c5676229
+size 639068

data/feats/23.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3829f5cfbeab2191b8732bbb9c0b2588236884d40463d2d9d9a0355f13739c5b
+size 639068

data/feats/24.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1e9e05493e450501aea9aa7b2f16fdce7618198790ea55d0f270272f10aabf94
+size 639068

data/feats/25.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9691048bfd0d0bc6e741564f1d33cb45054105e43b709072b645c650fdbe88e0
+size 639068

data/feats/26.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e9f2bdf9eec0114be38e77cc9bdf2fe181501069e26a51ba8cc39081182c806a
+size 639068

data/feats/27.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:efc8182b73dcfa2fcc0383fadf7c99f1e8841f21a4d356a4d57a4dbeb74188ce
+size 639068

data/feats/28.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c781fc06cdc504773822725a0cc0c833e752130a6ce5a52e0e7092f98f36db1a
+size 639068

data/feats/29.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ac185e131d3430267c3bf68b097f771ca09736e35120654d2a0734d1e4a93464
+size 639068

data/feats/3.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d05afa604a44460a6c6ad0ec33a5f8eeea2c344984de35efd2055956a0598389
+size 639065

data/feats/30.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:579cf8d4bb7bf2acce917b5cc109afc5511feb1ff7830a27373a0507001e5329
+size 639068

data/feats/31.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7815692babecedb80682afe8f8c6f6f4f377b8862490c82b9202680e10973caf
+size 639068

data/feats/32.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a7a2d4ffb0aa5019d6bb44e7921154e6db20510ac74c1bd45a429ef4a8b744e1
+size 639068

data/feats/33.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5d6465a5f845a6d00661c2f27da81e37761ea322e8da4813c8342272620c9319
+size 639068

data/feats/34.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5964f36395a4fbac147126adf4795fecd8eea24604df529b746f2115be51cfc7
+size 639068

data/feats/35.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6bafebef939cb64ff136e928d98ea6efd03d16396f9fedffae74c7fbcfc92f19
+size 639068

data/feats/36.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:55eabc3ef67a155f4722ad3064dfa65ad2d5b34b8e84084aa4a37215bc6173a4
+size 639068

data/feats/37.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d81924e4761c5b2ea4d7984fa0536ca3da1c97ff696bab4435cb95bb3a18db4e
+size 639068

data/feats/38.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d4b01d882d79d3de213abe45a441f3f5bb059d07bc725e63c5461ec6d5eef430
+size 639068

data/feats/39.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:110d36710a033a95e52dd7dabb150a41a3c793855d5dbfd6016d62f4b5fc43cd
+size 639068

data/feats/4.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c31158adb537f11bb73e7036be5df482e804f3198d74c9a84ba184f08f2acd1e
+size 639065

data/feats/40.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:366b353476790e157edba90bedb93649d145357477e8e1a42b661fec9c4f9f7f
+size 639068

data/feats/41.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2e7749780caaabf088b04388af24cf99391f8e4c9dd0a112456a518d74643827
+size 639068

data/feats/42.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a5db2fa43372581382aea1b8f8eb0642e3e18c32a431b85732a5021e6868ef8b
+size 639068

data/feats/43.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:405249437414400b5717b3ffb0e572aad382fd888a6f2964d75991a91a6157e5
+size 639068

data/feats/44.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fd7ef122fc0fdd1c6df65b9052b75bdf52cb9457673476aca80df5750cfd02e7
+size 639068

data/feats/45.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:860c9a0fdc60a18d7d54d31c98204fee976f6464a83bffd6d6acd5b6d2c45f4b
+size 639068

data/feats/46.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9df2d104b98d7d8467dfa3e1c58a067bc0d8abb4dd8b61cb0ac1d70c614d965c
+size 639068

data/feats/47.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:294149f4bdeda849bb80ff2acf1976c04c2b73225dbd42a27521b7569374ccfc
+size 639068

data/feats/48.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1a3bb1dd5c42af317e85f42d848f32d23c3f5315865b48e4b9ceddebf1e36ab5
+size 639068

data/feats/49.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b9920294263b7860c8428503b36e46adbd90fb1f3c6c04b61c36bf84e1a6edc4
+size 639068

data/feats/5.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:178b391bad95d3f103bd0c9bc91e335c3a6e160332c6c2621e68331860b1a1e3
+size 639065

data/feats/50.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9075f134133e24420eb65e08b607907965bc32b86466288f0e4ba5e3c189aa60
+size 639068