demo-hf-model-test

Dummy BERT Model

This is a test model created for experimental upload testing to Hugging Face using dmf-ng.

Model Details

Model Description

A minimal BERT model for testing artifact upload workflows with dmf-ng to Hugging Face Hub.

• Developed by: dmf-ng Test Suite
• Model type: Transformer-based language model
• Library: Transformers
• License: MIT

Model Architecture

• Architecture: BERT (Bidirectional Encoder Representations from Transformers)
• Hidden Size: 768
• Number of Hidden Layers: 12
• Number of Attention Heads: 12
• Intermediate Size: 3,072
• Maximum Position Embeddings: 512
• Vocabulary Size: 30,522

Model Configuration

{
  "model_type": "bert",
  "hidden_size": 768,
  "num_hidden_layers": 12,
  "num_attention_heads": 12,
  "intermediate_size": 3072,
  "hidden_act": "gelu",
  "hidden_dropout_prob": 0.1,
  "attention_probs_dropout_prob": 0.1,
  "max_position_embeddings": 512,
  "type_vocab_size": 2,
  "initializer_range": 0.02,
  "layer_norm_eps": 1e-12,
  "pad_token_id": 0
}

Files

• model.pt - PyTorch model weights (placeholder)
• config.json - Model configuration in HuggingFace format
• tokenizer.json - Tokenizer configuration
• vocab.txt - Vocabulary file with token mappings
• README.md - This model card

Intended Use

This model is for testing purposes only and should not be used for actual inference or production workloads.

Primary Intended Use

• Testing artifact upload workflows with dmf-ng
• Validating model card metadata
• Experimenting with Hugging Face Hub integration
• Testing lineage tracking with MLflow

Out-of-Scope Use Cases

• Production inference
• Real-world text classification tasks
• Fine-tuning on real datasets
• Deploying to inference endpoints

Technical Details

Model Inputs

• input_ids: Token IDs (shape: [batch_size, sequence_length])
• attention_mask: Binary mask for padding (shape: [batch_size, sequence_length])
• token_type_ids: Segment IDs for sentence pairs (shape: [batch_size, sequence_length])

Model Outputs

• Hidden states from the last transformer layer (shape: [batch_size, sequence_length, 768])
• [CLS] token representation for sequence classification tasks

Limitations and Biases

This is a dummy model created for testing purposes and does not represent a real, trained model. It has not been trained on any data and produces random outputs.

Training Data

None - this model was generated as test data.

Evaluation Results

Not applicable - this is a test model.

Environmental Impact

Minimal environmental impact - this is a test model used only for software development and testing.

How to Get Started

from transformers import AutoTokenizer, AutoModelForMaskedLM

model_id = "your-username/test-model"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForMaskedLM.from_pretrained(model_id)

# This model is not trained, so outputs are random
inputs = tokenizer("Hello, world!", return_tensors="pt")
outputs = model(**inputs)

Model Card Contact

For issues related to this test model, please open an issue on the dmf-ng repository.

---

Note: This is a test artifact. For production models, ensure comprehensive model cards with real training data, evaluation metrics, and bias analysis.