SentenceTransformer based on google/embeddinggemma-300m

Model Details

Model Description

• Model Type: Sentence Transformer
• Base model: google/embeddinggemma-300m
• Maximum Sequence Length: 2048 tokens
• Output Dimensionality: 768 dimensions
• Similarity Function: Cosine Similarity

Model Sources

• Documentation: Sentence Transformers Documentation
• Repository: Sentence Transformers on GitHub
• Hugging Face: Sentence Transformers on Hugging Face

Full Model Architecture

SentenceTransformer(
  (0): Transformer({'max_seq_length': 2048, 'do_lower_case': False, 'architecture': 'Gemma3TextModel'})
  (1): Pooling({'word_embedding_dimension': 768, 'pooling_mode_cls_token': False, 'pooling_mode_mean_tokens': True, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False, 'pooling_mode_weightedmean_tokens': False, 'pooling_mode_lasttoken': False, 'include_prompt': True})
  (2): Dense({'in_features': 768, 'out_features': 3072, 'bias': False, 'activation_function': 'torch.nn.modules.linear.Identity'})
  (3): Dense({'in_features': 3072, 'out_features': 768, 'bias': False, 'activation_function': 'torch.nn.modules.linear.Identity'})
  (4): Normalize()
)

Usage

Direct Usage (Sentence Transformers)

Training Details

Training Dataset

Unnamed Dataset

• Size: 10,000 training samples
• Columns: sentence_0 and sentence_1
• Approximate statistics based on the first 1000 samples:

  	sentence_0	sentence_1
  type	string	string
  details	min: 8 tokens mean: 64.43 tokens max: 577 tokens	min: 41 tokens mean: 329.7 tokens max: 1770 tokens

• Samples:

  sentence_0	sentence_1
  Explain the stock market indicators that analysts use to determine market trends and how they analyze companies within the market.	• Market indexes: Analysts track major stock market indexes like the S&P 500, Dow Jones Industrial Average, and Nasdaq Composite to get a sense of the overall market direction. When these indexes are rising consistently, it indicates a bull market, and when they are falling, it points to a bear market.[object Object][object Object]• Price-to-earnings (P/E) ratio: The P/E ratio compares a company's stock price to its earnings per share. Analysts use the P/E ratio to determine if a stock is overvalued or undervalued relative to its history and competitors. A higher P/E ratio suggests the stock is more expensive.[object Object][object Object]• Earnings: Analysts pay close attention to companies' quarterly and annual earnings reports. Beating or missing analyst earnings estimates can significantly impact a stock's price. Long-term earnings growth expectations also influence stock valuations.[object Object][object Object]• Economic indicators: Macroeconomic data like GDP growth, unemployment rates, and interest rates provide context for analysts to gauge overall demand in ...
  I am trying to create a calculator that allows users to specify asset classes (e.g. 500 in fund1, 500 in fund2, etc.) and the desired fraction of the total portfolio per asset class (e.g. 70% in fund1, 30% in fund2). The equation should determine how much to buy or sell per asset class (instead of move). I have the following formula for when only 1 fund changes, where $X_1$ is before, $X_2$ is after, and $dX$ is their difference. The same applies to the second fund $Y$. $a$ is the desired fraction of $X$. if $X_2 = a (X_2 + Y)$ and $X_2 = dX + X1$ is true, then $dX + X_1 = a (X_2 + Y)$. Rewriting that gave me: $$ dX = \frac{aX_1 - X_1 + aY}{1-a} $$ With $X_1 = 50$, $a = 0.70$, and $Y = 50$; $dX$ will be $66.5$. Now imagine a situation where users can enter multiple funds and the corresponding desired fraction at will. Now, whenever I calculate the amount to buy in one fund, the total changes, and any previous calculations need to be redone. It seems like I need a way to combine these ...	The normal way to do this is to calculate the current value $P$ of the complete portfolio. Then for each fund $i$ multiply it by the target fraction $a_i$ for that fund. This gives the target holding $X_i$ in fund $i$. Then $dX_i$ is the difference between between the target holding and the current holding in fund $i$. This allows you to calculate the needed changes for all funds at once.
  Calculate the total revenue generated by a business with a profit margin of 25%, fixed costs of $20,000 and variable costs of $4 per unit if they sell 10,000 units. Show your work and express your final answer as an exact dollar value.	* Given:[object Object] - Profit margin = 25%[object Object] - Fixed costs = $20,000 [object Object] - Variable costs = $4 per unit[object Object] - Units sold = 10,000[object Object]* Total variable costs = Variable cost per unit x Units sold[object Object] = $4 x 10,000[object Object] = $40,000 [object Object]* Total costs = Fixed costs + Variable costs[object Object] = $20,000 + $40,000[object Object] = $60,000[object Object]* Profit = Revenue - Total costs[object Object]* Since profit margin is 25%, we can say:[object Object] Profit/Revenue = 0.25[object Object] Profit = 0.25 * Revenue[object Object]* Plugging in the profit we found:[object Object] $15,000 = 0.25 * Revenue[object Object] Revenue = $15,000/0.25[object Object] Revenue = $60,000[object Object][object Object]So the total revenue generated is $60,000

• Loss: MultipleNegativesRankingLoss with these parameters:
  json

  1{ 2 "scale": 20.0, 3 "similarity_fct": "cos_sim", 4 "gather_across_devices": false 5}

Training Hyperparameters

Non-Default Hyperparameters

• per_device_train_batch_size: 2
• per_device_eval_batch_size: 2
• num_train_epochs: 1
• multi_dataset_batch_sampler: round_robin

All Hyperparameters

Training Logs

Framework Versions

• Python: 3.12.11
• Sentence Transformers: 5.1.0
• Transformers: 4.56.2
• PyTorch: 2.8.0+cu128
• Accelerate: 1.10.1
• Datasets: 4.0.0
• Tokenizers: 0.22.1

Citation

BibTeX

Sentence Transformers

MultipleNegativesRankingLoss