user@devops:~$ cat README.md
Hugging Face Text Generation
# Description
Text generation project with Hugging Face transformer models. Explores generative models like GPT-2, Llama 2 and Mistral-7B, and compares different decoding strategies: greedy search, beam search, sampling with temperature, top-k sampling and top-p (nucleus) sampling. Analyzes how each strategy affects generated text quality (fluency, diversity, coherence). Includes conditional generation examples (prompt engineering), parameter configuration (max_length, repetition_penalty, no_repeat_ngram_size) and perplexity evaluation as a quality metric.
# Key features
$ Generative models: GPT-2, Llama 2, Mistral-7B
$ Decoding strategies: greedy, beam search, sampling
$ Adjustable temperature, top-k and top-p (nucleus sampling)
$ Conditional generation with prompt engineering
$ Advanced config: repetition_penalty, no_repeat_ngram_size
$ Quality evaluation with perplexity metric
# Gallery
# Technologies used
