deploy: e9164c6

expconfig.html

@@ -571,7 +571,7 @@ <h2>Model Configurations<a class="headerlink" href="#model-configurations" title
 
 <dl class="py class">
 <dt class="sig sig-object py" id="realhf.ModelTrainEvalConfig">
-<em class="property"><span class="pre">class</span><span class="w"> </span></em><span class="sig-prename descclassname"><span class="pre">realhf.</span></span><span class="sig-name descname"><span class="pre">ModelTrainEvalConfig</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">type:</span> <span class="pre">~realhf.api.core.config.ModelFamily</span> <span class="pre">=</span> <span class="pre">llama-7</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">backend:</span> <span class="pre">str</span> <span class="pre">=</span> <span class="pre">'megatron'</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">path:</span> <span class="pre">str</span> <span class="pre">=</span> <span class="pre">''</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">lora:</span> <span class="pre">~realhf.api.quickstart.model.LoRAConfig</span> <span class="pre">|</span> <span class="pre">None</span> <span class="pre">=</span> <span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">gradient_checkpointing:</span> <span class="pre">bool</span> <span class="pre">=</span> <span class="pre">True</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">enable_fp16:</span> <span class="pre">bool</span> <span class="pre">=</span> <span class="pre">True</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">enable_bf16:</span> <span class="pre">bool</span> <span class="pre">=</span> <span class="pre">False</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">offload:</span> <span class="pre">bool</span> <span class="pre">=</span> <span class="pre">False</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">zero_stage:</span> <span class="pre">int</span> <span class="pre">=</span> <span class="pre">2</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">optimizer:</span> <span class="pre">~realhf.api.quickstart.model.OptimizerConfig</span> <span class="pre">|</span> <span class="pre">None</span> <span class="pre">=</span> <span class="pre">&lt;factory&gt;</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">init_critic_from_actor:</span> <span class="pre">bool</span> <span class="pre">=</span> <span class="pre">False</span></span></em><span class="sig-paren">)</span><a class="headerlink" href="#realhf.ModelTrainEvalConfig" title="Link to this definition">¶</a></dt>
+<em class="property"><span class="pre">class</span><span class="w"> </span></em><span class="sig-prename descclassname"><span class="pre">realhf.</span></span><span class="sig-name descname"><span class="pre">ModelTrainEvalConfig</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">type:</span> <span class="pre">~realhf.api.core.config.ModelFamily</span> <span class="pre">=</span> <span class="pre">llama-7</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">backend:</span> <span class="pre">str</span> <span class="pre">=</span> <span class="pre">'megatron'</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">path:</span> <span class="pre">str</span> <span class="pre">=</span> <span class="pre">''</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">lora:</span> <span class="pre">~realhf.api.quickstart.model.LoRAConfig</span> <span class="pre">|</span> <span class="pre">None</span> <span class="pre">=</span> <span class="pre">None</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">gradient_checkpointing:</span> <span class="pre">bool</span> <span class="pre">=</span> <span class="pre">True</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">enable_fp16:</span> <span class="pre">bool</span> <span class="pre">=</span> <span class="pre">True</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">enable_bf16:</span> <span class="pre">bool</span> <span class="pre">=</span> <span class="pre">False</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">offload:</span> <span class="pre">bool</span> <span class="pre">=</span> <span class="pre">False</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">zero_stage:</span> <span class="pre">int</span> <span class="pre">=</span> <span class="pre">2</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">optimizer:</span> <span class="pre">~realhf.api.quickstart.model.OptimizerConfig</span> <span class="pre">|</span> <span class="pre">None</span> <span class="pre">=</span> <span class="pre">&lt;factory&gt;</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">init_from_scratch:</span> <span class="pre">bool</span> <span class="pre">=</span> <span class="pre">False</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">init_critic_from_actor:</span> <span class="pre">bool</span> <span class="pre">=</span> <span class="pre">False</span></span></em><span class="sig-paren">)</span><a class="headerlink" href="#realhf.ModelTrainEvalConfig" title="Link to this definition">¶</a></dt>
 <dd><p>Runtime configuration for models (or LLMs) in ReaL.</p>
 <p>We use a customized model class instead of HuggingFace’s. This customized model has
 the following highlights:</p>