metadata
datasets:
- Skylion007/openwebtext
language:
- en
library_name: transformers
license: apache-2.0
metrics:
- perplexity
pipeline_tag: text-generation
LangFlow
LangFlow is a continuous diffusion language model that operates in embedding space. Unlike discrete diffusion models (MDLM, SEDD, DUO), LangFlow performs diffusion directly on continuous token embeddings, enabling smoother denoising dynamics. It is the first continuous DLM to rival discrete diffusion models on standard language modeling benchmarks like LM1B and OpenWebText.
- Paper: LangFlow: Continuous Diffusion Rivals Discrete in Language Modeling
- Code: GitHub Repository
- Project Blog: LangFlow Blog Post
Using LangFlow
To use the pre-trained model for text generation, use the following snippet:
from transformers import AutoModelForMaskedLM, AutoTokenizer
tokenizer = AutoTokenizer.from_pretrained('gpt2')
model = AutoModelForMaskedLM.from_pretrained('chumengl/langflow-owt', trust_remote_code=True)
# Generate samples
samples = model.generate_samples(num_samples=5, num_steps=128)
texts = tokenizer.batch_decode(samples, skip_special_tokens=True)
for text in texts:
print(text)
Model Details
- Architecture: DiT (Diffusion Transformer) backbone with adaptive layer normalization
- Context Length: 1024 tokens
- Parameters: ~130M non-embedding parameters (similar to GPT-2 medium)
- Training: 1M steps on OpenWebText corpus
- Tokenizer: GPT-2 tokenizer (50,257 vocab size)
Model Card Contact
Chumeng Liang (chumengl@illinois.edu)