Spaces:
Running
Running
File size: 1,787 Bytes
fb8f2dc ae51d62 b4f3263 ae51d62 b4f3263 933b7b6 b4f3263 ae51d62 b4f3263 1c1651d ae51d62 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 |
import os
from transformers import AutoTokenizer, AutoModelForCausalLM
import torch
from torch.nn.functional import cosine_similarity
from collections import Counter
import numpy as np
from device_manager import DeviceManager
class Gemma2BDependencies:
_instance = None
def __new__(cls):
if cls._instance is None:
cls._instance = super(Gemma2BDependencies, cls).__new__(cls)
hf_token = os.environ.get('HUGGINGFACE_TOKEN', None)
cls._instance.tokenizer = AutoTokenizer.from_pretrained("google/gemma-2b", token=hf_token)
cls._instance.model = AutoModelForCausalLM.from_pretrained("google/gemma-2b", token=hf_token)
cls._instance.device = DeviceManager()
cls._instance.model.to(cls._instance.device)
return cls._instance
def calculate_perplexity(self, text: str):
inputs = self.tokenizer(text, return_tensors="pt",
truncation=True, max_length=1024)
inputs = {k: v.to(self.device) for k, v in inputs.items()}
# Calculate the model's output
with torch.no_grad():
outputs = self.model(**inputs, labels=inputs["input_ids"])
loss = outputs.loss
perplexity = torch.exp(loss)
return perplexity.item()
def calculate_burstiness(self, text: str):
tokens = self.tokenizer.encode(text, add_special_tokens=False)
# Count token frequencies
frequency_counts = list(Counter(tokens).values())
# Calculate variance and mean of frequencies
variance = np.var(frequency_counts)
mean = np.mean(frequency_counts)
# Compute Variance-to-Mean Ratio (VMR) for burstiness
vmr = variance / mean if mean > 0 else 0
return vmr
|