--- language: en thumbnail: https://github.com/borisdayma/huggingtweets/blob/master/img/logo_share.png?raw=true tags: - huggingtweets widget: - text: "My dream is" --- <link rel="stylesheet" href="https://unpkg.com/@tailwindcss/typography@0.2.x/dist/typography.min.css"> <style> @media (prefers-color-scheme: dark) { .prose { color: #E2E8F0 !important; } .prose h2, .prose h3, .prose a, .prose thead { color: #F7FAFC !important; } } </style> <section class='prose'> <div> <div style="width: 132px; height:132px; border-radius: 50%; background-size: cover; background-image: url('http://pbs.twimg.com/profile_images/517689241812627456/pyBGyEo__400x400.jpeg')"> </div> <div style="margin-top: 8px; font-size: 19px; font-weight: 800">Lutz Büch 🤖 AI Bot </div> <div style="font-size: 15px; color: #657786">@geilehirnbude bot</div> </div> I was made with [huggingtweets](https://github.com/borisdayma/huggingtweets). Create your own bot based on your favorite user with [the demo](https://colab.research.google.com/github/borisdayma/huggingtweets/blob/master/huggingtweets-demo.ipynb)! ## How does it work? The model uses the following pipeline.  To understand how the model was developed, check the [W&B report](https://app.wandb.ai/wandb/huggingtweets/reports/HuggingTweets-Train-a-model-to-generate-tweets--VmlldzoxMTY5MjI). ## Training data The model was trained on [@geilehirnbude's tweets](https://twitter.com/geilehirnbude). <table style='border-width:0'> <thead style='border-width:0'> <tr style='border-width:0 0 1px 0; border-color: #CBD5E0'> <th style='border-width:0'>Data</th> <th style='border-width:0'>Quantity</th> </tr> </thead> <tbody style='border-width:0'> <tr style='border-width:0 0 1px 0; border-color: #E2E8F0'> <td style='border-width:0'>Tweets downloaded</td> <td style='border-width:0'>3116</td> </tr> <tr style='border-width:0 0 1px 0; border-color: #E2E8F0'> <td style='border-width:0'>Retweets</td> <td style='border-width:0'>2906</td> </tr> <tr style='border-width:0 0 1px 0; border-color: #E2E8F0'> <td style='border-width:0'>Short tweets</td> <td style='border-width:0'>53</td> </tr> <tr style='border-width:0'> <td style='border-width:0'>Tweets kept</td> <td style='border-width:0'>157</td> </tr> </tbody> </table> [Explore the data](https://app.wandb.ai/wandb/huggingtweets/runs/wt8lffrr/artifacts), which is tracked with [W&B artifacts](https://docs.wandb.com/artifacts) at every step of the pipeline. ## Training procedure The model is based on a pre-trained [GPT-2](https://huggingface.co/gpt2) which is fine-tuned on @geilehirnbude's tweets. Hyperparameters and metrics are recorded in the [W&B training run](https://app.wandb.ai/wandb/huggingtweets/runs/1u8augcw) for full transparency and reproducibility. At the end of training, [the final model](https://app.wandb.ai/wandb/huggingtweets/runs/1u8augcw/artifacts) is logged and versioned. ## Intended uses & limitations ### How to use You can use this model directly with a pipeline for text generation: <pre><code><span style="color:#03A9F4">from</span> transformers <span style="color:#03A9F4">import</span> pipeline generator = pipeline(<span style="color:#FF9800">'text-generation'</span>, model=<span style="color:#FF9800">'huggingtweets/geilehirnbude'</span>) generator(<span style="color:#FF9800">"My dream is"</span>, num_return_sequences=<span style="color:#8BC34A">5</span>)</code></pre> ### Limitations and bias The model suffers from [the same limitations and bias as GPT-2](https://huggingface.co/gpt2#limitations-and-bias). In addition, the data present in the user's tweets further affects the text generated by the model. ## About *Built by Boris Dayma* </section> [](https://twitter.com/borisdayma) <section class='prose'> For more details, visit the project repository. </section> [](https://github.com/borisdayma/huggingtweets)