Upload 63 files

data/analogy.txt

@@ -0,0 +1,28 @@
+рубль россия сша доллар
+рубль россия европа евро
+москва россия украина киев
+москва россия сша вашингтон
+москва россия япония токио
+москва россия германия берлин
+москва россия фрг берлин
+москва россия сербия белград
+москва россия китай пекин
+москва россия венгрия будапешт
+москва россия белоруссия минск
+москва россия бельгия брюссель
+москва россия таджикистан душанбе
+москва россия франция париж
+москва россия испания мадрид
+москва россия финляндия хельсинки
+отец мужчина женщина мать
+сын мужчина женщина дочь
+мальчик мужчина женщина девочка
+актёр мужчина женщина актриса
+родственник мужчина женщина родственница
+артист мужчина женщина артистка
+пассажир мужчина женщина пассажирка
+пенсионер мужчина женщина пенсионерка
+певец мужчина женщина певица
+писатель мужчина женщина писательница
+российский россия украина украинский
+российский россия сша американский

data/antonyms.txt

@@ -0,0 +1,20 @@
+вопрос ответ
+мужчина женщина
+мальчик девочка
+полный частичный
+зима лето
+чёрный белый
+плюс минус
+молодой пожилой
+дорогой дешёвый
+рождение смерть
+день ночь
+личный общественный
+близкий дальний
+выше ниже
+свадьба развод
+родиться умереть
+экспорт импорт
+запад восток
+полный пустой
+верхний нижний

data/axis.txt

@@ -0,0 +1,4 @@
+мужчина женщина
+молодой пожилой
+россия украина
+запад восток

data/nearest_neighbors.txt

@@ -0,0 +1,6 @@
+мужчина
+женщина
+россия
+человек
+конфликт
+контакт

data/synonyms.txt

@@ -0,0 +1,11 @@
+страна государство
+найти обнаружить
+сегмент часть фрагмент доля
+разный различный
+спорт физкультура
+союз коалиция
+просто легко
+врач доктор
+питаться есть
+алкоголь спиртное
+письмо сообщение

main.py

@@ -0,0 +1,452 @@
+import fasttext
+import streamlit as st
+import numpy as np
+import pandas as pd
+from gensim.models import Word2Vec
+from sklearn.metrics.pairwise import cosine_similarity
+import plotly.express as px
+import plotly.graph_objects as go
+from collections import Counter
+import os
+import glob
+
+
+class UnifiedVectorModel:
+    def __init__(self, backend_model, model_type="w2v"):
+        self.model = backend_model
+        self.model_type = model_type.lower()
+
+        if self.model_type == "w2v":
+            self.wv = backend_model.wv
+            self.key_to_index = self.wv.key_to_index
+            self.vector_size = self.wv.vector_size
+            self._words = set(self.wv.key_to_index.keys())
+
+        elif self.model_type == "ft":
+            # Для fasttext-wheel
+            self.key_to_index = {word: i for i, word in enumerate(backend_model.get_words())}
+            self.vector_size = backend_model.get_dimension()
+            self._words = set(self.key_to_index.keys())
+        else:
+            raise ValueError("model_type must be 'w2v' or 'ft'")
+
+    def __contains__(self, word):
+        return word in self._words
+
+    def __getitem__(self, word):
+        if self.model_type == "w2v":
+            return self.wv[word]
+        elif self.model_type == "ft":
+            return self.model.get_word_vector(word)
+
+    def most_similar(self, positive=None, negative=None, topn=10):
+        from sklearn.metrics.pairwise import cosine_similarity
+
+        if not positive:
+            positive = []
+        if not negative:
+            negative = []
+
+        try:
+            if self.model_type == "w2v":
+                return self.wv.most_similar(positive=positive, negative=negative, topn=topn)
+
+            elif self.model_type == "ft":
+                vec = np.zeros(self.vector_size)
+                for w in positive:
+                    if w in self:
+                        vec += self[w]
+                    else:
+                        continue
+                for w in negative:
+                    if w in self:
+                        vec -= self[w]
+                    else:
+                        continue
+
+                if np.allclose(vec, 0):
+                    return []
+
+                words = list(self._words)
+                vectors = np.array([self[w] for w in words])
+
+                sims = cosine_similarity([vec], vectors)[0]
+                best = np.argsort(sims)[::-1][:topn + len(positive) + len(negative)]
+
+                result = []
+                for i in best:
+                    word = words[i]
+                    if word not in positive and word not in negative:
+                        result.append((word, float(sims[i])))
+                        if len(result) >= topn:
+                            break
+                return result
+
+        except Exception as e:
+            print(f"Error in most_similar: {e}")
+            return []
+
+    def similar_by_vector(self, vector, topn=10):
+        from sklearn.metrics.pairwise import cosine_similarity
+
+        words = list(self._words)
+        vectors = np.array([self[w] for w in words])
+        sims = cosine_similarity([vector], vectors)[0]
+        best = np.argsort(sims)[::-1][:topn]
+
+        return [(words[i], float(sims[i])) for i in best]
+
+    def get_words(self):
+        return list(self._words)
+
+    @property
+    def vectors(self):
+        if not hasattr(self, '_cached_vectors'):
+            words = list(self._words)
+            self._cached_words = words
+            self._cached_vectors = np.array([self[w] for w in words])
+        return self._cached_vectors
+
+    @property
+    def index_to_key(self):
+        if not hasattr(self, '_index_to_key'):
+            self._index_to_key = list(self._words)
+        return self._index_to_key
+
+
+@st.cache_resource
+def load_model(model_path):
+    try:
+        if model_path.endswith(".model"):
+            raw_model = Word2Vec.load(model_path)
+            current_model = UnifiedVectorModel(raw_model, model_type="w2v")
+
+        elif model_path.endswith(".bin"):
+            raw_model = fasttext.load_model(model_path)
+            current_model = UnifiedVectorModel(raw_model, model_type="ft")
+        else:
+            raise ValueError(f"wrong path format")
+        return current_model
+    except Exception as e:
+        st.error(f"error loading model {model_path}: {e}")
+        return None
+
+
+MODELS_DIR = "models"
+
+if not os.path.exists(MODELS_DIR):
+    st.error(f"Folder `{MODELS_DIR}` not found.")
+    st.stop()
+
+model_files = []
+for ext in ["*.bin", "*.model", "*.vec"]:
+    model_files.extend(glob.glob(os.path.join(MODELS_DIR, ext)))
+model_files = [f for f in model_files if os.path.isfile(f)]
+model_names = [os.path.basename(f) for f in model_files]
+
+if len(model_names) == 0:
+    st.error(f"No models in folder `{MODELS_DIR}` (.bin, .model, .vec).")
+    st.info("Supported formats: Word2Vec (binary/text), FastText.")
+    st.stop()
+
+selected_model_name = st.sidebar.selectbox(
+    "Choose pretrained model",
+    model_names
+)
+
+selected_model_path = os.path.join(MODELS_DIR, selected_model_name)
+
+st.sidebar.info(f"loading: `{selected_model_name}`")
+
+model = load_model(selected_model_path)
+
+if model is None:
+    st.stop()
+else:
+    st.sidebar.success(f"Model '{selected_model_name}' loaded")
+    st.sidebar.write(f"Voc size: {len(model.key_to_index):,}")
+    st.sidebar.write(f"Vector size: {model.vector_size}")
+
+def analogy_accuracy(model, file_name):
+    right = 0
+    count = 0
+    results = []
+    with open(file_name, encoding='utf-8') as file:
+        for line in file:
+            words = line.strip().split()
+            if len(words) != 4:
+                continue
+            try:
+                most_similar = model.most_similar(positive=[words[0], words[2]], negative=[words[1]], topn=10)
+                predicted = [x[0] for x in most_similar]
+                correct = words[3]
+                if correct in predicted:
+                    rank = predicted.index(correct) + 1
+                    right += 1
+                else:
+                    rank = None
+                count += 1
+                results.append({
+                    "query": f"{words[0]} - {words[1]} + {words[2]}",
+                    "target": correct,
+                    "predicted": predicted[0],
+                    "rank": rank,
+                    "in_top10": bool(rank)
+                })
+            except KeyError as e:
+                continue
+    accuracy = right / count if count > 0 else 0
+    return accuracy, results
+
+
+def avg_similarity(model, file_name):
+    res = []
+    with open(file_name, encoding='utf-8') as file:
+        for line in file:
+            words = line.strip().split()
+            try:
+                vectors = [model[word] for word in words]
+            except KeyError:
+                continue
+            sims = cosine_similarity(vectors)
+            for i in range(len(words) - 1):
+                for j in range(i + 1, len(words)):
+                    res.append(sims[i][j])
+    return sum(res) / len(res) if res else 0
+
+
+def projection(word_vec, axis):
+    axis_norm = axis / np.linalg.norm(axis)
+    return np.dot(word_vec, axis_norm)
+
+
+def get_projection_row(model, axis):
+    words = list(model.key_to_index.keys())
+    projections = [(word, projection(model[word], axis)) for word in words]
+    projections = sorted(projections, key=lambda x: x[1])
+    return projections
+
+
+st.title("Vector embeddings")
+
+tab1, tab2, tab3, tab4, tab5 = st.tabs([
+    "Vector ariphmetics",
+    "Semantic consistency",
+    "Semantic axis",
+    "Distribution analysis",
+    "Report"
+])
+
+with tab1:
+    st.header("Vector ariphmetics")
+    expr = st.text_input("Insert expression", value="рубль - россия + сша")
+
+    if st.button("Compute"):
+        words = expr.replace('+', ' + ').replace('-', ' - ').split()
+        positive, negative = [], []
+        current = 'pos'
+
+        for w in words:
+            if w == '+':
+                current = 'pos'
+            elif w == '-':
+                current = 'neg'
+            else:
+                (positive if current == 'pos' else negative).append(w)
+
+        missing = [w for w in positive + negative if w not in model]
+        if missing:
+            st.warning(f"Words not found in voc: {', '.join(missing)}")
+            st.stop()
+
+        try:
+            similar = model.most_similar(
+                positive=positive,
+                negative=negative,
+                topn=10
+            )
+
+            st.write("### Result:")
+            result_words = [f"{w} ({s:.3f})" for w, s in similar]
+            st.write("Nearest words: " + ", ".join(result_words))
+
+            st.write("### In-between steps")
+
+            cum_vec = np.zeros(model.vector_size)
+
+            steps_data = []
+
+            for i in range(len(positive)):
+                cum_vec += model[w]
+                nearest = model.most_similar(positive=positive[:i + 1], topn=1)
+                steps_data.append({
+                    "step": f"+ {positive[i]}",
+                    "nearest word": nearest[0][0],
+                    "similarity": nearest[0][1]
+                })
+
+            for i in range(len(negative)):
+                cum_vec -= model[w]
+                nearest = model.most_similar(positive=positive, negative=negative[:i + 1], topn=1)
+                steps_data.append({
+                    "step": f"- {negative[i]}",
+                    "nearest word": nearest[0][0],
+                    "similarity": nearest[0][1]
+                })
+
+            df_steps = pd.DataFrame(steps_data)
+            st.dataframe(df_steps[["step", "nearest word", "similarity"]])
+
+            result_word = similar[0][0]
+            fig = px.scatter(
+                x=[cum_vec[0]], y=[cum_vec[1]],
+                text=[result_word],
+                title="Result (first 2 components)"
+            )
+            fig.update_traces(textposition='top center', marker=dict(size=12, color='red'))
+            st.plotly_chart(fig)
+
+        except Exception as e:
+            st.error(f"Error computing: {e}")
+
+with tab2:
+    st.header("Similarity calculator")
+    col1, col2 = st.columns(2)
+    with col1:
+        word1 = st.text_input("word 1", value="мужчина")
+    with col2:
+        word2 = st.text_input("word 2", value="женщина")
+
+    if st.button("Compute similarity"):
+        try:
+            v1, v2 = model[word1], model[word2]
+            sim = cosine_similarity([v1], [v2])[0][0]
+            st.metric("Cosine similarity", f"{sim:.4f}")
+
+            st.write("### Nearest neighbors graph")
+            neighbors = model.most_similar(word1, topn=5) + model.most_similar(word2, topn=5)
+            nodes = list(set([word1, word2] + [n[0] for n in neighbors]))
+            edges = [(word1, n[0]) for n in model.most_similar(word1, topn=5)] + \
+                    [(word2, n[0]) for n in model.most_similar(word2, topn=5)]
+
+            G = go.Figure()
+            pos = np.random.rand(len(nodes), 2) * 2 - 1
+            node_x = pos[:, 0]
+            node_y = pos[:, 1]
+
+            for edge in edges:
+                x0, y0 = pos[nodes.index(edge[0])]
+                x1, y1 = pos[nodes.index(edge[1])]
+                G.add_trace(go.Scatter(x=[x0, x1], y=[y0, y1], mode='lines', line=dict(width=1, color='gray'), showlegend=False))
+
+            G.add_trace(go.Scatter(x=node_x, y=node_y, mode='text+markers',
+                                   marker=dict(size=10, color='lightblue'),
+                                   text=nodes, textposition="top center"))
+            G.update_layout(title="Semantic links graph", showlegend=False)
+            st.plotly_chart(G)
+
+        except KeyError as e:
+            st.error(f"Word not found: {e}")
+
+with tab3:
+    st.header("Semantic axis projection")
+    col1, col2 = st.columns(2)
+    with col1:
+        pos_axis = st.text_input("positive", value="мужчина")
+    with col2:
+        neg_axis = st.text_input("negative", value="женщина")
+
+    if st.button("Build axis"):
+        try:
+            pos_vec = model[pos_axis]
+            neg_vec = model[neg_axis]
+            axis = pos_vec - neg_vec
+
+            projections = get_projection_row(model, axis)
+            top_pos = projections[-10:][::-1]
+            top_neg = projections[:10]
+
+            st.write(f"Axis: **{pos_axis} – {neg_axis}**")
+            st.write("### Top 10 positive:")
+            st.write(", ".join([f"{w} ({p:.3f})" for w, p in top_pos]))
+
+            st.write("### Top 10 negative:")
+            st.write(", ".join([f"{w} ({p:.3f})" for w, p in top_neg]))
+
+            df_proj = pd.DataFrame(top_pos + top_neg, columns=["word", "projection"])
+            fig = px.bar(df_proj, x="projection", y="word", orientation='h', title=f"Projection on axis: {pos_axis}–{neg_axis}")
+            st.plotly_chart(fig)
+
+        except KeyError as e:
+            st.error(f"Error: {e}")
+
+with tab4:
+    st.header("Distance distribution analysis")
+    all_vectors = model.vectors
+    sample = all_vectors[np.random.choice(all_vectors.shape[0], 1000, replace=False)]
+
+    dists = cosine_similarity(sample)
+    np.fill_diagonal(dists, 0)
+    flat_dists = dists.flatten()
+    flat_dists = flat_dists[flat_dists > 0]
+
+    fig = px.histogram(flat_dists, nbins=50, title="Cosine similarity distribution between random words")
+    st.plotly_chart(fig)
+
+    st.metric("Mean similarity", f"{np.mean(flat_dists):.3f}")
+    st.metric("Std deviation", f"{np.std(flat_dists):.3f}")
+
+with tab5:
+    st.header("Report")
+
+    st.subheader("1. Analogy rate")
+    analogies_file = "data/analogy.txt"
+    if os.path.exists(analogies_file):
+        acc, results = analogy_accuracy(model, analogies_file)
+        st.metric("Analogy accuracy (in top 10)", f"{acc:.2%}")
+        st.dataframe(pd.DataFrame(results))
+    else:
+        st.warning("File `analogy.txt` not found.")
+
+    st.subheader("2. Average synonyms similarity")
+    sim_file = "data/synonyms.txt"
+    if os.path.exists(sim_file):
+        avg_sim = avg_similarity(model, sim_file)
+        st.metric("Average similarity", f"{avg_sim:.4f}")
+    else:
+        st.warning("File `similarity_words.txt` not found.")
+
+    st.subheader("3. Average antonyms similarity")
+    sim_file = "data/antonyms.txt"
+    if os.path.exists(sim_file):
+        avg_sim = avg_similarity(model, sim_file)
+        st.metric("Average similarity", f"{avg_sim:.4f}")
+    else:
+        st.warning("File `similarity_words.txt` not found.")
+
+    st.subheader("4. Heatmap for nearest words")
+    query_words = st.text_input("Enter words", value="мужчина женщина мальчик девочка").split()
+    if st.button("Build heatmap"):
+        try:
+            vectors = [model[w] for w in query_words]
+            sims = cosine_similarity(vectors)
+            fig = px.imshow(sims, x=query_words, y=query_words, color_continuous_scale="Blues", title="Similarity heatmap")
+            st.plotly_chart(fig)
+        except KeyError as e:
+            st.error(f"Error: {e}")
+
+    st.subheader("5. 2D projection")
+    sample_words = st.text_input("Input words", value="мужчина женщина мальчик девочка")
+    word_list = sample_words.split()
+    if st.button("Show clusters"):
+        try:
+            from sklearn.manifold import TSNE
+            vectors = np.array([model[w] for w in word_list])
+            tsne = TSNE(n_components=2, perplexity=len(vectors) - 1, random_state=42)
+            embedded = tsne.fit_transform(vectors)
+
+            fig = px.scatter(x=embedded[:, 0], y=embedded[:, 1], text=word_list, title="words projection")
+            fig.update_traces(textposition='top center')
+            st.plotly_chart(fig)
+        except KeyError as e:
+            st.error(f"Word not found: {e}")

models/w2v_v100_w10_m10_sg0.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2c35ee5be232fce591fcf28be6fde61bd4a90d75fc7262d7303adaf0b32806b0
+size 3176561

models/w2v_v100_w10_m10_sg1.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:95b361b695fcb9e78d2209b1ebd5ca2e8b3b024479e7085329d2f0b6444fc2fe
+size 3176559

models/w2v_v100_w10_m5_sg0.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7cf858300a2a007c6a56718fb112b4320009e75d967884b2d74a9c9567dcc0c7
+size 5158115

models/w2v_v100_w10_m5_sg1.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c1fe1d49f365e7fd8acc11f2988e9cb8ba04061badf74337c9b13618735908d1
+size 5158113

models/w2v_v100_w10_m8_sg0.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3c5bd023602c41ed6c6d512953246e23f7913848db8cf8495587c780681f7b8c
+size 3790868

models/w2v_v100_w10_m8_sg1.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3e4475e87311f026e925ab1ad9f816021f807570727f3b30256249032e4d85e7
+size 3790868

models/w2v_v100_w5_m10_sg0.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d3cd4410cd1063c041beed75d5e9a02c039aefe95a38157d2adc268014167584
+size 3176557

models/w2v_v100_w5_m10_sg1.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6898fee7f31bb81a2c877cc3be3725eb4428ffe1650f876dc48737930aa5da1b
+size 3176557

models/w2v_v100_w5_m5_sg0.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c550467186f9723753553e16288ee05e7e22b6fa55b5ede806d76c20d1badff6
+size 5158113

models/w2v_v100_w5_m5_sg1.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:439758d8fe5b0efd096f70c6bf34c8f8f756d2ff3fd5be33c827b9fafe5087c6
+size 5158113

models/w2v_v100_w5_m8_sg0.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4a908b5b41ae2bc3cbe1b0688f9ed06422dc8e0cb3779c88a3129e0377364037
+size 3790866

models/w2v_v100_w5_m8_sg1.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3e698df12b684fb4bdc4ab2476ee29c07ddd3b51ca28b8b8a174e38d198f5fe5
+size 3790868

models/w2v_v100_w8_m10_sg0.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0a888d8a90f39928cee7539cc682a6c37d9139322d1e8d0a4f804cebba68fefd
+size 3176559

models/w2v_v100_w8_m10_sg1.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8fb4add03702502739b77b70d172eceba480375874898829e90e56789286c2ee
+size 3176557

models/w2v_v100_w8_m5_sg0.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9543f75d5113261d16ae3b13ed1e9904613be04dee0cdfb6c7b8f65841d422d3
+size 5158111

models/w2v_v100_w8_m5_sg1.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:baaac6ffa83930c5f8a39ded46a61524d0afe07f37139a61e64a83e9f9a6cfca
+size 5158111

models/w2v_v100_w8_m8_sg0.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:64064be1788ab5e1b97c78dd0fb5159bb9ca6e2e886ab64f7e46d1c81392a7bc
+size 3790866

models/w2v_v100_w8_m8_sg1.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b6f366a3495c390add6716f65eb0c29c8779d2407f6e004abb3396a1254811c6
+size 3790866

models/w2v_v200_w10_m10_sg0.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b83369f0f0ac975d2433d16c068e51eb1c21281eeadc2ccc99b95b73d7bb4bd8
+size 6201360

models/w2v_v200_w10_m10_sg1.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0a345190f56c1b26a52c9a123e6836f8be99dd5cfbc4f55779c5f9db9e38e5af
+size 6201361

models/w2v_v200_w10_m5_sg0.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d79ad8709c9f181ad4ae2176c9e7e7b1d6ad0fd4ec1883c5de59c4bae0fda27c
+size 10071716

models/w2v_v200_w10_m5_sg1.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7e1c09904ca66331a49fabfe78ef2936b579e02da4e04679e1e0d29c59e74bc2
+size 10071713

models/w2v_v200_w10_m8_sg0.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8b0f536e8576e94fd3c815ce8b6f969768eea4b6dce1df1df71a6ca0af1e2c04
+size 7401271

models/w2v_v200_w10_m8_sg1.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bddae5328e51ee1d6a37a4b1be0a8792a4a0b86f992ee1b0c8cc56c513eee136
+size 7401268

models/w2v_v200_w5_m10_sg0.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fa99e8e7ea983a3044ebe686f74dd875ba948cde92d251915784c11a8faec586
+size 6201359

models/w2v_v200_w5_m10_sg1.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2e55bc410d00ad641b52122d2d9cb436bb10af7129c7bfa36892c043bdced086
+size 6201359

models/w2v_v200_w5_m5_sg0.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:46ceb9f48cf980b11561c64c6e98f82be85afdbd9b3ad5d03bcee8a1d876c149
+size 10071712

models/w2v_v200_w5_m5_sg1.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:717cd37fe973d2a76f615d5a093cd42be93d081ef2cea770aae4b53a1ff44d0d
+size 10071711

models/w2v_v200_w5_m8_sg0.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9d4a1eb433cbad61cd37743ec5846ca06938c1ad9f9ef8b27d6ec1740dd8d401
+size 7401268

models/w2v_v200_w5_m8_sg1.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:71174e2974a2f12b92f1a14a13b046eb26ee9a28cd0136ec5211ed151ccf7ff5
+size 7401268

models/w2v_v200_w8_m10_sg0.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e48303d6345ec48580eb29913656335842f6db4cb0e524566a997ea10a105c09
+size 6201357

models/w2v_v200_w8_m10_sg1.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3ec3208cc8936180db231957865d412230531c76cc0891b8acb5ac51e4208c71
+size 6201357

models/w2v_v200_w8_m5_sg0.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:35189d2c165137323d2fcb7b58604903f5e1236a6be432d5a7378fb1c2773cf1
+size 10071714

models/w2v_v200_w8_m5_sg1.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:06f920ef0f10fb0b3995c1e35bb15027faae7d936ba6a54a114e5ac00a524a26
+size 10071713

models/w2v_v200_w8_m8_sg0.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:21c865d402a86cdbb35ee3f803428c77c3cb97845c2f7307ae941a59c31bce45
+size 7401269

models/w2v_v200_w8_m8_sg1.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:247bf90c6c2b9c5305a7f9453af64ad655b5a26856f3634c7a023c2c2992759c
+size 7401266

models/w2v_v300_w10_m10_sg0.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:69df3844797167fcdeaf8753f0f274873e17b3f78f7041a506909c4ac02d7e39
+size 9226167

models/w2v_v300_w10_m10_sg1.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bb1492cdd7b3b6242b6d06307437182e30e0757173783c8934985b89a347f8a6
+size 9226166

models/w2v_v300_w10_m5_sg0.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a72fada5eac351a9c5787a63111d145de5043f6ad686f0307575289730921544
+size 14985321

models/w2v_v300_w10_m5_sg1.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fe98dac1a0f0669e48f6c8943d1380e4d8f841c304929a95825c88f8ed41facc
+size 14985318

models/w2v_v300_w10_m8_sg0.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:643b9eef53b5e85f73a70b155e85a0137ccb248bb20708945e699b244e803b31
+size 11011674