نمایش دو بعدی embeddings

 1import pandas as pd
 2import tiktoken
 3
 4# load & inspect dataset
 5input_datapath = "data/fine_food_reviews_1k.csv"  # to save space, we provide a pre-filtered dataset
 6df = pd.read_csv(input_datapath, index_col=0)
 7df = df[["Time", "ProductId", "UserId", "Score", "Summary", "Text"]]
 8df = df.dropna()
 9df["combined"] = (
10    "Title: " + df.Summary.str.strip() + "; Content: " + df.Text.str.strip()
11)
12df.head(2)

 1
 2embedding_model = "text-embedding-3-small"
 3max_tokens = 8000  # the maximum for text-embedding-3-small is 8191
 4embedding_encoding = "cl100k_base"
 5
 6# subsample to 1k most recent reviews and remove samples that are too long
 7top_n = 1000
 8df = df.sort_values("Time").tail(top_n * 2)  # first cut to first 2k entries, assuming less than half will be filtered out
 9df.drop("Time", axis=1, inplace=True)
10
11encoding = tiktoken.get_encoding(embedding_encoding)
12
13# omit reviews that are too long to embed
14df["n_tokens"] = df.combined.apply(lambda x: len(encoding.encode(x)))
15df = df[df.n_tokens <= max_tokens].tail(top_n)

 1
 2from openai import OpenAI # for calling the OpenAI API
 3import os
 4
 5client = OpenAI(
 6    api_key=os.environ.get(("GILAS_API_KEY", "<کلید API خود را اینجا بسازید https://dashboard.gilas.io/apiKey>")), 
 7    base_url="https://api.gilas.io/v1/" # Gilas APIs
 8)
 9
10def get_embedding(query)
11    query_embedding_response = client.embeddings.create(
12        model=embedding_model,
13        input=query,
14    )
15    return query_embedding_response.data[0].embedding
16
17
18# This may take a few minutes
19df["embedding"] = df.combined.apply(lambda x: get_embedding(x))
20df.to_csv("data/fine_food_reviews_with_embeddings_1k.csv")

 1import pandas as pd
 2from sklearn.manifold import TSNE
 3import numpy as np
 4from ast import literal_eval
 5
 6# Load the embeddings
 7datafile_path = "data/fine_food_reviews_with_embeddings_1k.csv"
 8df = pd.read_csv(datafile_path)
 9
10# Convert to a list of lists of floats
11matrix = np.array(df.embedding.apply(literal_eval).to_list())
12
13# Create a t-SNE model and transform the data
14tsne = TSNE(n_components=2, perplexity=15, random_state=42, init='random', learning_rate=200)
15vis_dims = tsne.fit_transform(matrix)
16vis_dims.shape

 1import matplotlib.pyplot as plt
 2import matplotlib
 3import numpy as np
 4
 5colors = ["red", "darkorange", "gold", "turquoise", "darkgreen"]
 6x = [x for x,y in vis_dims]
 7y = [y for x,y in vis_dims]
 8color_indices = df.Score.values - 1
 9
10colormap = matplotlib.colors.ListedColormap(colors)
11plt.scatter(x, y, c=color_indices, cmap=colormap, alpha=0.3)
12for score in [0,1,2,3,4]:
13    avg_x = np.array(x)[df.Score-1==score].mean()
14    avg_y = np.array(y)[df.Score-1==score].mean()
15    color = colors[score]
16    plt.scatter(avg_x, avg_y, marker='x', color=color, s=100)
17
18plt.title("Amazon ratings visualized in language using t-SNE")