All public logs

Combined display of all available logs of Marovi AI. You can narrow down the view by selecting a log type, the username (case-sensitive), or the affected page (also case-sensitive).

00:20, 9 September 2024 Felipefelixarias talk contribs created page Translations:Diffusion Models Are Real-Time Game Engines/26/zh (Created page with "=== 3.2 训练生成扩散模型 ===")
00:20, 9 September 2024 Felipefelixarias talk contribs created page Translations:Diffusion Models Are Real-Time Game Engines/25/zh (Created page with "我们在整个训练过程中记录了代理的训练轨迹，其中涵盖了不同技能水平的游戏。这组记录的轨迹构成了我们的<math>\mathcal{T}_{agent}</math>数据集，用于训练生成模型（见第[https://arxiv.org/html/2408.14837v1#S3.SS2 3.2]节）。")
00:20, 9 September 2024 Felipefelixarias talk contribs created page Translations:Diffusion Models Are Real-Time Game Engines/24/zh (Created page with "我们的最终目标是让人类玩家与我们的仿真进行互动。为此，第[https://arxiv.org/html/2408.14837v1#S2 2]节中的策略<math>\pi</math>即为“人类游戏策略”。由于我们无法直接大规模地从中取样，因此我们首先通过教一个自动代理来玩游戏，以此来近似人类游戏。与典型的强化学习设置不同，该设置旨在最大化游戏得分，我们的目标是生成与人类游戏类似的训练数据，或...")
00:20, 9 September 2024 Felipefelixarias talk contribs created page Translations:Diffusion Models Are Real-Time Game Engines/23/zh (Created page with "=== 3.1 通过代理进行数据收集 ===")
00:20, 9 September 2024 Felipefelixarias talk contribs created page Translations:Diffusion Models Are Real-Time Game Engines/22/zh (Created page with "center|thumb|900x900px|图3：GameNGen方法概览。为了简洁起见，省略了v预测的详细信息。")
00:20, 9 September 2024 Felipefelixarias talk contribs created page Translations:Diffusion Models Are Real-Time Game Engines/21/zh (Created page with "GameNGen（发音为“游戏引擎”）是一个生成扩散模型，它能够在第[https://arxiv.org/html/2408.14837v1#S2 2]节的设置下学习模拟游戏。为了收集该模型的训练数据，我们首先使用教师强制目标训练一个独立的模型与环境进行交互。这两个模型（代理和生成模型）依次进行训练。在训练过程中，代理的全部行为和观察语料 <math>\mathcal{T}_{agent}</math> 被保留下来，并在第二...")
00:20, 9 September 2024 Felipefelixarias talk contribs created page Translations:Diffusion Models Are Real-Time Game Engines/20/zh (Created page with "== 3 GameNGen ==")
00:19, 9 September 2024 Felipefelixarias talk contribs created page Translations:Diffusion Models Are Real-Time Game Engines/19/zh (Created page with "我们总是使用教师强迫目标来训练我们的生成模型。给定一个模拟分布函数 <math>q</math>，可以通过自回归地采样观测值来模拟环境 <math>\mathcal{E}</math>。")
00:19, 9 September 2024 Felipefelixarias talk contribs created page Translations:Diffusion Models Are Real-Time Game Engines/18/zh (Created page with "给定输入交互环境 <math>\mathcal{E}</math> 和初始状态 <math>s_{0} \in \mathcal{S}</math>，一个“交互世界模拟”是一个“模拟分布函数” <math>q \left( o_{n} \,|\, \{o_{< n}, a_{\leq n}\} \right), \; o_{i} \in \mathcal{O}, \; a_{i} \in \mathcal{A}</math>。给定观测值之间的距离度量 <math>D: \mathcal{O} \times \mathcal{O} \rightarrow \mathbb{R}</math>，一个“策略”，即给定过去动作和观测的代理动作分布 <math>...")
00:19, 9 September 2024 Felipefelixarias talk contribs created page Translations:Diffusion Models Are Real-Time Game Engines/17/zh (Created page with "例如，在游戏 DOOM 中，<math>\mathcal{S}</math> 是程序的动态内存内容，<math>\mathcal{O}</math> 是渲染的屏幕像素，<math>V</math> 是游戏的渲染逻辑，<math>\mathcal{A}</math> 是按键和鼠标移动的集合，而 <math>p</math> 是基于玩家输入的程序逻辑（包括任何潜在的非确定性）。")
00:19, 9 September 2024 Felipefelixarias talk contribs created page Translations:Diffusion Models Are Real-Time Game Engines/16/zh (Created page with "一个''交互环境''<math>\mathcal{E}</math>由一个潜在状态空间<math>\mathcal{S}</math>、一个潜在空间的部分投影空间<math>\mathcal{O}</math>、一个部分投影函数<math>V: \mathcal{S} \rightarrow \mathcal{O}</math>、一组动作<math>\mathcal{A}</math>，以及一个转移概率函数<math>p \left( s^{\prime} \,|\, a, s \right)</math>，使得<math>s, s^{\prime} \in \mathcal{S}, a\in \mathcal{A}</math>。")
00:19, 9 September 2024 Felipefelixarias talk contribs created page Translations:Diffusion Models Are Real-Time Game Engines/15/zh (Created page with "== 2 互动世界仿真 ==")
00:19, 9 September 2024 Felipefelixarias talk contribs created page Translations:Diffusion Models Are Real-Time Game Engines/14/zh (Created page with "center|thumb|800x800px|图 2：GameNGen 与之前最先进的 DOOM 仿真的比较")
00:19, 9 September 2024 Felipefelixarias talk contribs created page Translations:Diffusion Models Are Real-Time Game Engines/13/zh (Created page with "GameNGen 回答了在通往游戏引擎新范式的道路上一个重要的问题，即游戏可以自动生成，就像近年来神经模型生成图像和视频一样。仍然存在关键问题，例如如何训练这些神经游戏引擎，以及如何有效地创建游戏，包括如何最佳地利用人类输入。尽管如此，我们对这种新范式的可能性感到非常兴奋。")
00:19, 9 September 2024 Felipefelixarias talk contribs created page Translations:Diffusion Models Are Real-Time Game Engines/12/zh (Created page with "在这项工作中，我们证明答案是肯定的。具体来说，我们展示了一款复杂的视频游戏——标志性游戏《DOOM》，可以在神经网络（开放式 Stable Diffusion v1.4 的增强版（Rombach 等人，[https://arxiv.org/html/2408.14837v1#bib.bib26 2022]））上实时运行，同时获得与原始游戏相当的视觉质量。尽管这不是精确仿真，该神经模型能够执行复杂的游戏状态更新，例如统计生命值和弹...")
00:18, 9 September 2024 Felipefelixarias talk contribs created page Translations:Diffusion Models Are Real-Time Game Engines/11/zh (Created page with "一个实时运行的神经模型是否能够以高质量模拟复杂的游戏？")
00:18, 9 September 2024 Felipefelixarias talk contribs created page Translations:Diffusion Models Are Real-Time Game Engines/10/zh (Created page with "有几项重要研究（Ha & Schmidhuber，[https://arxiv.org/html/2408.14837v1#bib.bib10 2018]；Kim 等人，[https://arxiv.org/html/2408.14837v1#bib.bib16 2020]；Bruce 等人，[https://arxiv.org/html/2408.14837v1#bib.bib7 2024]）（见第[https://arxiv.org/html/2408.14837v1#S6 6]节）使用神经模型来模拟交互式视频游戏。然而，这些方法大多在模拟游戏的复杂性、仿真速度、长时间的稳定性或视觉质量等方面存在局限性...")
00:18, 9 September 2024 Felipefelixarias talk contribs created page Translations:Diffusion Models Are Real-Time Game Engines/9/zh (Created page with "近年来，生成模型在根据文本或图像等多模态输入生成图像和视频方面取得了重大进展。在这一浪潮的前沿，扩散模型成为非语言媒体生成的事实标准，如 Dall-E（Ramesh 等人，[https://arxiv.org/html/2408.14837v1#bib.bib25 2022]）、Stable Diffusion（Rombach 等人，[https://arxiv.org/html/2408.14837v1#bib.bib26 2022]）和 Sora（Brooks 等人，[https://arxiv.org/html/2408.14837v1#bib.bib6 2024]）。乍一看，...")
00:18, 9 September 2024 Felipefelixarias talk contribs created page Translations:Diffusion Models Are Real-Time Game Engines/8/zh (Created page with "计算机游戏是围绕以下“游戏循环”手动制作的软件系统：(1) 收集用户输入，(2) 更新游戏状态，(3) 将其渲染为屏幕像素。这个游戏循环以很高的帧率运行，为玩家营造出一个交互式虚拟世界的假象。这种游戏循环通常在标准计算机上运行，尽管也有许多在定制硬件上运行游戏的惊人尝试（例如，标志性游戏《毁灭战士》曾在烤面包机、微波炉、跑步机、照...")
00:18, 9 September 2024 Felipefelixarias talk contribs created page Translations:Diffusion Models Are Real-Time Game Engines/7/zh (Created page with "== 1 介绍 ==")
00:18, 9 September 2024 Felipefelixarias talk contribs created page Translations:Diffusion Models Are Real-Time Game Engines/6/zh (Created page with "center|thumb|800x800px|图 1：一名玩家正在 '''GameNGen''' 上以 20 FPS 的速度游玩 DOOM。请参见 [https://gamengen.github.io/ https://gamengen.github.io] 获取演示视频。")
00:18, 9 September 2024 Felipefelixarias talk contribs created page Translations:Diffusion Models Are Real-Time Game Engines/5/zh (Created page with "我们介绍了''GameNGen''，这是第一个完全由神经模型驱动的游戏引擎，能够在长轨迹上与复杂环境进行高质量的实时交互。GameNGen 可以在单个 TPU 上以每秒超过 20 帧的速度交互模拟经典游戏 DOOM。下一帧预测的 PSNR 为 29.4，与有损 JPEG 压缩相当。在区分游戏短片和模拟片段方面，人类评分员的表现仅略好于随机概率。GameNGen 的训练分为两个阶段：(1) 一个强化学...")
00:18, 9 September 2024 Felipefelixarias talk contribs created page Translations:Diffusion Models Are Real-Time Game Engines/4/zh (Created page with "====== 摘要 ======")
00:04, 9 September 2024 Felipefelixarias talk contribs created page Diffusion Models Are Real-Time Game Engines/zh (Created page with "'''项目网站：''' [https://gamengen.github.io/ https://gamengen.github.io]")
00:04, 9 September 2024 Felipefelixarias talk contribs created page Translations:Diffusion Models Are Real-Time Game Engines/3/zh (Created page with "'''项目网站：''' [https://gamengen.github.io/ https://gamengen.github.io]")
00:03, 9 September 2024 Felipefelixarias talk contribs created page Translations:Diffusion Models Are Real-Time Game Engines/2/zh (Created page with "'''ArXiv链接：''' [https://arxiv.org/abs/2408.14837 https://arxiv.org/abs/2408.14837]")
00:03, 9 September 2024 Felipefelixarias talk contribs created page Translations:Diffusion Models Are Real-Time Game Engines/1/zh (Created page with "'''作者：''' Dani Valevski（谷歌研究）、Yaniv Leviathan（谷歌研究）、Moab Arar（特拉维夫大学）、Shlomi Fruchter（谷歌 DeepMind）")
02:16, 8 September 2024 Felipefelixarias talk contribs marked Diffusion Models Are Real-Time Game Engines for translation
02:07, 8 September 2024 Felipefelixarias talk contribs marked Diffusion Models Are Real-Time Game Engines for translation
02:07, 8 September 2024 Felipefelixarias talk contribs marked Diffusion Models Are Real-Time Game Engines for translation
18:25, 7 September 2024 Felipefelixarias talk contribs marked Welcome for translation
06:57, 7 September 2024 Felipefelixarias talk contribs marked Welcome for translation
06:51, 7 September 2024 Felipefelixarias talk contribs marked Welcome for translation
06:44, 7 September 2024 Felipefelixarias talk contribs created page File:FelipeFelixArias2024.jpg
06:44, 7 September 2024 Felipefelixarias talk contribs uploaded File:FelipeFelixArias2024.jpg
06:34, 7 September 2024 Felipefelixarias talk contribs created page Translations:Diffusion Models Are Real-Time Game Engines/45/es (Created page with "Observamos que es sencillo aumentar aún más la tasa de generación de imágenes de manera sustancial al paralelizar la generación de varios fotogramas en hardware adicional, similar a la técnica clásica de Nvidia SLI Alternate Frame Rendering (AFR). Al igual que con AFR, la tasa real de simulación no aumentaría ni se reduciría el retardo de entrada.")
06:34, 7 September 2024 Felipefelixarias talk contribs created page Translations:Diffusion Models Are Real-Time Game Engines/44/es (Created page with "Dado que observamos degradación al utilizar un solo paso de muestreo, también experimentamos con la destilación de modelos de manera similar a (Yin et al., [https://arxiv.org/html/2408.14837v1#bib.bib39 2024]; Wang et al., [https://arxiv.org/html/2408.14837v1#bib.bib36 2023]) en el entorno de un solo paso. La destilación ayuda sustancialmente en este caso (permitiéndonos alcanzar los 50 FPS como se mencionó antes), pero aún conlleva cierto coste para la calidad de...")
06:34, 7 September 2024 Felipefelixarias talk contribs created page Translations:Diffusion Models Are Real-Time Game Engines/43/es (Created page with "El uso de solo 4 pasos de denoising resulta en un costo total de U-Net de 40 ms (y un costo total de inferencia de 50 ms, incluyendo el auto-codificador) o 20 fotogramas por segundo. Nuestra hipótesis es que el impacto insignificante en la calidad con pocos pasos en nuestro caso se debe a una combinación de: (1) un espacio de imágenes restringido, y (2) un fuerte condicionamiento por los fotogramas anteriores.")
06:34, 7 September 2024 Felipefelixarias talk contribs created page Translations:Diffusion Models Are Real-Time Game Engines/42/es (Created page with "Durante la inferencia, necesitamos ejecutar tanto el denoizador U-Net (durante una serie de pasos) como el autocodificador. En nuestra configuración de hardware (un TPU-v5), tanto un único paso del denoizador como una evaluación del autocodificador tardan 10 ms. Si ejecutáramos nuestro modelo con un único paso del denoizador, la latencia total mínima posible en nuestra configuración sería de 20 ms por fotograma, o 50 fotogramas por segundo. Normalmente, los model...")
06:34, 7 September 2024 Felipefelixarias talk contribs created page Translations:Diffusion Models Are Real-Time Game Engines/41/es (Created page with "==== 3.3.2 Pasos de Muestreo del Denoizador ====")
06:33, 7 September 2024 Felipefelixarias talk contribs created page Translations:Diffusion Models Are Real-Time Game Engines/40/es (Created page with "También experimentamos con la generación de 4 muestras en paralelo y la combinación de los resultados, con la esperanza de evitar que se aceptaran predicciones extremas poco frecuentes y reducir la acumulación de errores. Probamos tanto promediar las muestras como elegir la muestra más cercana a la mediana. El promediado funcionó ligeramente peor que un solo fotograma, y elegir la más cercana a la mediana funcionó sólo marginalmente mejor. Dado que ambos aumenta...")
06:33, 7 September 2024 Felipefelixarias talk contribs created page Translations:Diffusion Models Are Real-Time Game Engines/39/es (Created page with "Utilizamos el muestreo DDIM (Song et al., [https://arxiv.org/html/2408.14837v1#bib.bib34 2022]). Empleamos la Guía Sin Clasificador (Ho & Salimans, [https://arxiv.org/html/2408.14837v1#bib.bib12 2022]) solo para la condición de observaciones pasadas <math>o_{< n}</math>. No encontramos guía para la condición de acciones pasadas <math>a_{< n}</math> que mejorara la calidad. La ponderación que utilizamos es relativamente pequeña (1.5), ya que ponderaciones mayores cr...")
06:33, 7 September 2024 Felipefelixarias talk contribs created page Translations:Diffusion Models Are Real-Time Game Engines/38/es (Created page with "==== 3.3.1 Configuración ====")
06:32, 7 September 2024 Felipefelixarias talk contribs created page Translations:Diffusion Models Are Real-Time Game Engines/36/es (Created page with "El autocodificador preentrenado de Stable Diffusion v1.4, que comprime parches de 8x8 píxeles en 4 canales latentes, produce artefactos significativos al predecir los fotogramas del juego, lo que afecta a los pequeños detalles y, en particular, a la barra inferior del HUD («heads-up display»). Para aprovechar el conocimiento preentrenado y mejorar al mismo tiempo la calidad de la imagen, entrenamos solo el decodificador del autocodificador latente utilizando una pér...")
06:32, 7 September 2024 Felipefelixarias talk contribs created page Translations:Diffusion Models Are Real-Time Game Engines/35/es (Created page with "==== 3.2.2 Ajuste Fino del Decodificador Latente ====")
06:32, 7 September 2024 Felipefelixarias talk contribs created page Translations:Diffusion Models Are Real-Time Game Engines/33/es (Created page with "El cambio de dominio entre el entrenamiento con el objetivo de forzado por el maestro y el muestreo autorregresivo lleva a la acumulación de errores y a una rápida degradación de la calidad de la muestra, como se demuestra en la figura [https://arxiv.org/html/2408.14837v1#S3.F4 4]. Para evitar esta divergencia debida a la aplicación autorregresiva del modelo, corrompemos los fotogramas de contexto añadiendo una cantidad variable de ruido gaussiano a los fotogramas c...")
06:32, 7 September 2024 Felipefelixarias talk contribs created page Translations:Diffusion Models Are Real-Time Game Engines/32/es (Created page with "==== 3.2.1 Mitigación de la Deriva Autorregresiva Utilizando la Augmentación de Ruido ====")
06:31, 7 September 2024 Felipefelixarias talk contribs created page Translations:Diffusion Models Are Real-Time Game Engines/30/es (Created page with "<math>\mathcal{L} = {{\mathbb{E}}_{t,\epsilon,T}\left\lbrack {\|{{v{(\epsilon,x_{0},t)}} - {v_{\theta^{\prime}}{(x_{t},t,{\{{\phi{(o_{i < n})}}\}},{\{{A_{emb}{(a_{i < n})}}\}})}}}\|}_{2}^{2} \right\rbrack}</math> (1)")
06:31, 7 September 2024 Felipefelixarias talk contribs created page Translations:Diffusion Models Are Real-Time Game Engines/27/es (Created page with "Ahora entrenamos un modelo generativo de difusión condicionado a las trayectorias del agente <math>\mathcal{T}_{agent}</math> (acciones y observaciones) recopiladas durante la etapa anterior.")
06:31, 7 September 2024 Felipefelixarias talk contribs created page Translations:Diffusion Models Are Real-Time Game Engines/24/es (Created page with "Nuestro objetivo final es que los jugadores humanos interactúen con nuestra simulación. Para ello, la política <math>\pi</math> como en la sección [https://arxiv.org/html/2408.14837v1#S2 2] es la del ''juego humano''. Dado que no podemos tomar muestras de eso directamente a gran escala, comenzamos por aproximarlo enseñando a un agente automático a jugar. A diferencia de una configuración típica de RL que intenta maximizar la puntuación del juego, nuestro objetiv...")