Buckets:

hf-doc-build/doc-dev / transformers /pr_42227 /en /trainer_customize.html
HuggingFaceDocBuilder's picture
download
raw
23.1 kB
<meta charset="utf-8" /><meta name="hf:doc:metadata" content="{&quot;title&quot;:&quot;Subclassing Trainer methods&quot;,&quot;local&quot;:&quot;subclassing-trainer-methods&quot;,&quot;sections&quot;:[{&quot;title&quot;:&quot;get_train_dataloader&quot;,&quot;local&quot;:&quot;gettraindataloader&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;compute_loss&quot;,&quot;local&quot;:&quot;computeloss&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;Next steps&quot;,&quot;local&quot;:&quot;next-steps&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2}],&quot;depth&quot;:1}">
<link href="/docs/transformers/pr_42227/en/_app/immutable/assets/0.e3b0c442.css" rel="modulepreload">
<link rel="modulepreload" href="/docs/transformers/pr_42227/en/_app/immutable/entry/start.a067d14a.js">
<link rel="modulepreload" href="/docs/transformers/pr_42227/en/_app/immutable/chunks/scheduler.31fdf58d.js">
<link rel="modulepreload" href="/docs/transformers/pr_42227/en/_app/immutable/chunks/singletons.303ebed4.js">
<link rel="modulepreload" href="/docs/transformers/pr_42227/en/_app/immutable/chunks/index.252883d5.js">
<link rel="modulepreload" href="/docs/transformers/pr_42227/en/_app/immutable/chunks/paths.afba2d50.js">
<link rel="modulepreload" href="/docs/transformers/pr_42227/en/_app/immutable/entry/app.95d76f53.js">
<link rel="modulepreload" href="/docs/transformers/pr_42227/en/_app/immutable/chunks/preload-helper.21822fa8.js">
<link rel="modulepreload" href="/docs/transformers/pr_42227/en/_app/immutable/chunks/index.2f76fdf0.js">
<link rel="modulepreload" href="/docs/transformers/pr_42227/en/_app/immutable/nodes/0.cb4700b8.js">
<link rel="modulepreload" href="/docs/transformers/pr_42227/en/_app/immutable/chunks/each.e59479a4.js">
<link rel="modulepreload" href="/docs/transformers/pr_42227/en/_app/immutable/nodes/672.6ee80ec5.js">
<link rel="modulepreload" href="/docs/transformers/pr_42227/en/_app/immutable/chunks/CopyLLMTxtMenu.bb5a0eb6.js">
<link rel="modulepreload" href="/docs/transformers/pr_42227/en/_app/immutable/chunks/MermaidChart.svelte_svelte_type_style_lang.d0fb4fa3.js">
<link rel="modulepreload" href="/docs/transformers/pr_42227/en/_app/immutable/chunks/IconCopy.ac192424.js">
<link rel="modulepreload" href="/docs/transformers/pr_42227/en/_app/immutable/chunks/CodeBlock.ab12f8e1.js"><!-- HEAD_svelte-u9bgzb_START --><meta name="hf:doc:metadata" content="{&quot;title&quot;:&quot;Subclassing Trainer methods&quot;,&quot;local&quot;:&quot;subclassing-trainer-methods&quot;,&quot;sections&quot;:[{&quot;title&quot;:&quot;get_train_dataloader&quot;,&quot;local&quot;:&quot;gettraindataloader&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;compute_loss&quot;,&quot;local&quot;:&quot;computeloss&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;Next steps&quot;,&quot;local&quot;:&quot;next-steps&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2}],&quot;depth&quot;:1}"><!-- HEAD_svelte-u9bgzb_END --> <p></p> <div class="items-center shrink-0 min-w-[100px] max-sm:min-w-[50px] justify-end ml-auto flex" style="float: right; margin-left: 10px; display: inline-flex; position: relative; z-index: 10;"><div class="inline-flex rounded-md max-sm:rounded-sm"><button class="inline-flex items-center gap-1 h-7 max-sm:h-7 px-2 max-sm:px-1.5 text-sm font-medium text-gray-800 border border-r-0 rounded-l-md max-sm:rounded-l-sm border-gray-200 bg-white hover:shadow-inner dark:border-gray-850 dark:bg-gray-950 dark:text-gray-200 dark:hover:bg-gray-800" aria-live="polite"><span class="inline-flex items-center justify-center rounded-md p-0.5 max-sm:p-0 hover:text-gray-800 dark:hover:text-gray-200"><svg class="sm:size-3.5 size-3" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg></span> <span>Copy page</span></button> <button class="inline-flex items-center justify-center w-6 max-sm:w-5 h-7 max-sm:h-7 disabled:pointer-events-none text-sm text-gray-500 hover:text-gray-700 dark:hover:text-white rounded-r-md max-sm:rounded-r-sm border border-l transition border-gray-200 bg-white hover:shadow-inner dark:border-gray-850 dark:bg-gray-950 dark:text-gray-200 dark:hover:bg-gray-800" aria-haspopup="menu" aria-expanded="false" aria-label="Open copy menu"><svg class="transition-transform text-gray-400 overflow-visible sm:size-3.5 size-3 rotate-0" width="1em" height="1em" viewBox="0 0 12 7" fill="none" xmlns="http://www.w3.org/2000/svg"><path d="M1 1L6 6L11 1" stroke="currentColor"></path></svg></button></div> </div> <h1 class="relative group"><a id="subclassing-trainer-methods" class="header-link block pr-1.5 text-lg no-hover:hidden with-hover:absolute with-hover:p-1.5 with-hover:opacity-0 with-hover:group-hover:opacity-100 with-hover:right-full" href="#subclassing-trainer-methods"><span><svg class="" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" aria-hidden="true" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 256 256"><path d="M167.594 88.393a8.001 8.001 0 0 1 0 11.314l-67.882 67.882a8 8 0 1 1-11.314-11.315l67.882-67.881a8.003 8.003 0 0 1 11.314 0zm-28.287 84.86l-28.284 28.284a40 40 0 0 1-56.567-56.567l28.284-28.284a8 8 0 0 0-11.315-11.315l-28.284 28.284a56 56 0 0 0 79.196 79.197l28.285-28.285a8 8 0 1 0-11.315-11.314zM212.852 43.14a56.002 56.002 0 0 0-79.196 0l-28.284 28.284a8 8 0 1 0 11.314 11.314l28.284-28.284a40 40 0 0 1 56.568 56.567l-28.285 28.285a8 8 0 0 0 11.315 11.314l28.284-28.284a56.065 56.065 0 0 0 0-79.196z" fill="currentColor"></path></svg></span></a> <span>Subclassing Trainer methods</span></h1> <p data-svelte-h="svelte-1y5sy5c">Subclass <a href="/docs/transformers/pr_42227/en/main_classes/trainer#transformers.Trainer">Trainer</a> methods to change training behavior without rewriting the entire loop. Subclassing modifies the <em>training loop</em>, for example the forward pass or loss computation.</p> <p data-svelte-h="svelte-26mwkh">Before subclassing, consider whether you need to change <em>what</em> <a href="/docs/transformers/pr_42227/en/main_classes/trainer#transformers.Trainer">Trainer</a> computes or <em>when</em> and <em>whether</em> it acts. For timing and conditional logic, use a <a href="./trainer_callbacks">Callback</a> instead. Callbacks control when things happen (logging, evaluation, early stopping) and subclassing changes what happens (loss computation, data loading, optimization).</p> <blockquote class="note" data-svelte-h="svelte-18v8qvs"><p>See the <a href="/docs/transformers/pr_42227/en/main_classes/trainer#transformers.Trainer">Trainer</a> API docs for a complete list of methods you can subclass. Private methods (prefixed with <code>_</code>) like <code>_save_checkpoint</code> or <code>_evaluate</code> can also be overridden, but these may change without notice.</p></blockquote> <h2 class="relative group"><a id="gettraindataloader" class="header-link block pr-1.5 text-lg no-hover:hidden with-hover:absolute with-hover:p-1.5 with-hover:opacity-0 with-hover:group-hover:opacity-100 with-hover:right-full" href="#gettraindataloader"><span><svg class="" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" aria-hidden="true" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 256 256"><path d="M167.594 88.393a8.001 8.001 0 0 1 0 11.314l-67.882 67.882a8 8 0 1 1-11.314-11.315l67.882-67.881a8.003 8.003 0 0 1 11.314 0zm-28.287 84.86l-28.284 28.284a40 40 0 0 1-56.567-56.567l28.284-28.284a8 8 0 0 0-11.315-11.315l-28.284 28.284a56 56 0 0 0 79.196 79.197l28.285-28.285a8 8 0 1 0-11.315-11.314zM212.852 43.14a56.002 56.002 0 0 0-79.196 0l-28.284 28.284a8 8 0 1 0 11.314 11.314l28.284-28.284a40 40 0 0 1 56.568 56.567l-28.285 28.285a8 8 0 0 0 11.315 11.314l28.284-28.284a56.065 56.065 0 0 0 0-79.196z" fill="currentColor"></path></svg></span></a> <span>get_train_dataloader</span></h2> <p data-svelte-h="svelte-1e7t2f6">The standard <a href="/docs/transformers/pr_42227/en/main_classes/trainer#transformers.Trainer.get_train_dataloader">get_train_dataloader()</a> method loads one batch, trains on it, discards it, and loads the next batch.</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-keyword">def</span> <span class="hljs-title function_">get_train_dataloader</span>(<span class="hljs-params">self</span>):
<span class="hljs-keyword">return</span> self._get_dataloader(
batch_size=self._train_batch_size,
...
)<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-15et51k"><a href="https://huggingface.co/docs/trl/en/grpo_trainer" rel="nofollow">GRPO</a> is an online reinforcement learning algorithm that generates completions before training on them. Generating completions every step is expensive because it’s autoregressive. A 512-token completion requires ~512 sequential forward passes compared to one forward pass for a training step. <a href="https://huggingface.co/docs/trl/main/en/gspo_token#trl.GRPOTrainer" rel="nofollow">GRPOTrainer</a> subclasses <a href="/docs/transformers/pr_42227/en/main_classes/trainer#transformers.Trainer.get_train_dataloader">get_train_dataloader()</a> to batch generation across multiple steps.</p> <p data-svelte-h="svelte-891dg7"><code>trl.GRPOTrainer.get_train_dataloader</code> loads <em>batches</em> of generation prompts for multiple training steps at once by multiplying batch size by a <code>steps_per_generation</code> argument. If <code>train_batch_size=4</code> and <code>steps_per_generation=8</code>, the dataloader produces batches of 32, cutting generation cost by 8x.</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-keyword">def</span> <span class="hljs-title function_">get_train_dataloader</span>(<span class="hljs-params">self</span>):
dataloader_params = {
<span class="hljs-string">&quot;batch_size&quot;</span>: self._train_batch_size * self.args.steps_per_generation, <span class="hljs-comment"># this is the only change</span>
...
}<!-- HTML_TAG_END --></pre></div> <h2 class="relative group"><a id="computeloss" class="header-link block pr-1.5 text-lg no-hover:hidden with-hover:absolute with-hover:p-1.5 with-hover:opacity-0 with-hover:group-hover:opacity-100 with-hover:right-full" href="#computeloss"><span><svg class="" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" aria-hidden="true" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 256 256"><path d="M167.594 88.393a8.001 8.001 0 0 1 0 11.314l-67.882 67.882a8 8 0 1 1-11.314-11.315l67.882-67.881a8.003 8.003 0 0 1 11.314 0zm-28.287 84.86l-28.284 28.284a40 40 0 0 1-56.567-56.567l28.284-28.284a8 8 0 0 0-11.315-11.315l-28.284 28.284a56 56 0 0 0 79.196 79.197l28.285-28.285a8 8 0 1 0-11.315-11.314zM212.852 43.14a56.002 56.002 0 0 0-79.196 0l-28.284 28.284a8 8 0 1 0 11.314 11.314l28.284-28.284a40 40 0 0 1 56.568 56.567l-28.285 28.285a8 8 0 0 0 11.315 11.314l28.284-28.284a56.065 56.065 0 0 0 0-79.196z" fill="currentColor"></path></svg></span></a> <span>compute_loss</span></h2> <p data-svelte-h="svelte-d70q61"><a href="/docs/transformers/pr_42227/en/main_classes/trainer#transformers.Trainer.compute_loss">compute_loss()</a> returns the cross-entropy loss calculated by the model.</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-keyword">def</span> <span class="hljs-title function_">compute_loss</span>(<span class="hljs-params">self, model, inputs, return_outputs=<span class="hljs-literal">False</span>, num_items_in_batch=<span class="hljs-literal">None</span></span>):
...
outputs = model(**inputs)
...
loss = outputs[<span class="hljs-string">&quot;loss&quot;</span>] <span class="hljs-comment"># get loss from model</span>
<span class="hljs-keyword">return</span> (loss, outputs) <span class="hljs-keyword">if</span> return_outputs <span class="hljs-keyword">else</span> loss<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-xwuxf4"><a href="https://huggingface.co/docs/trl/en/dpo_trainer" rel="nofollow">DPO</a> measures how strongly the policy model prefers a chosen response over a rejected one, relative to a reference model. <a href="https://huggingface.co/docs/trl/main/en/bema_for_reference_model#trl.DPOTrainer" rel="nofollow">DPOTrainer</a> subclasses <a href="/docs/transformers/pr_42227/en/main_classes/trainer#transformers.Trainer.compute_loss">compute_loss()</a> because the loss computation differs from standard cross-entropy in several ways:</p> <ul data-svelte-h="svelte-1enfna7"><li>the model never sees labels; it only returns logits for DPO to calculate log-probs from</li> <li>chosen and rejected responses are concatenated</li> <li>a reference model calculates its own log-probs</li> <li>the loss is a function of <code>π_chosen</code>, <code>π_rejected</code>, <code>π_ref_chosen</code>, <code>π_ref_rejected</code></li></ul> <p data-svelte-h="svelte-nisqje">None of the above fits the standard <a href="/docs/transformers/pr_42227/en/main_classes/trainer#transformers.Trainer.compute_loss">Trainer.compute_loss()</a> method.</p> <div class="code-block relative "><div class="absolute top-2.5 right-4"><button class="inline-flex items-center relative text-sm focus:text-green-500 cursor-pointer focus:outline-none transition duration-200 ease-in-out opacity-0 mx-0.5 text-gray-600 " title="code excerpt" type="button"><svg class="" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M28,10V28H10V10H28m0-2H10a2,2,0,0,0-2,2V28a2,2,0,0,0,2,2H28a2,2,0,0,0,2-2V10a2,2,0,0,0-2-2Z" transform="translate(0)"></path><path d="M4,18H2V4A2,2,0,0,1,4,2H18V4H4Z" transform="translate(0)"></path><rect fill="none" width="32" height="32"></rect></svg> <div class="absolute pointer-events-none transition-opacity bg-black text-white py-1 px-2 leading-tight rounded font-normal shadow left-1/2 top-full transform -translate-x-1/2 translate-y-2 opacity-0"><div class="absolute bottom-full left-1/2 transform -translate-x-1/2 w-0 h-0 border-black border-4 border-t-0" style="border-left-color: transparent; border-right-color: transparent; "></div> Copied</div></button></div> <pre class=""><!-- HTML_TAG_START --><span class="hljs-keyword">def</span> <span class="hljs-title function_">compute_loss</span>(<span class="hljs-params">
self,
model: PreTrainedModel | nn.Module,
inputs: <span class="hljs-built_in">dict</span>[<span class="hljs-built_in">str</span>, torch.Tensor | <span class="hljs-type">Any</span>],
return_outputs=<span class="hljs-literal">False</span>,
num_items_in_batch=<span class="hljs-literal">None</span>,
</span>) -&gt; torch.Tensor | <span class="hljs-built_in">tuple</span>[torch.Tensor, <span class="hljs-built_in">dict</span>[<span class="hljs-built_in">str</span>, <span class="hljs-built_in">float</span>]]:
...
outputs = model(**inputs)
logits = outputs.logits
logps = get_logps(logits, inputs)
chosen_logps, rejected_logps = logps.chunk(<span class="hljs-number">2</span>, dim=<span class="hljs-number">0</span>) <span class="hljs-comment"># batch is [chosen, rejected]</span>
ref_logits = self.ref_model(**inputs).logits
ref_logps = get_logps(ref_logits, inputs)
ref_chosen_logps, ref_rejected_logps = ref_logps.chunk(<span class="hljs-number">2</span>, dim=<span class="hljs-number">0</span>) <span class="hljs-comment"># batch is [chosen, rejected]</span>
chosen_scores = chosen_logps - ref_chosen_logps
rejected_scores = rejected_logps - ref_rejected_logps
per_sequence_loss = -F.logsigmoid(self.beta * chosen_scores - rejected_scores)
loss = per_sequence_loss.mean()
<span class="hljs-keyword">return</span> (loss, outputs) <span class="hljs-keyword">if</span> return_outputs <span class="hljs-keyword">else</span> loss<!-- HTML_TAG_END --></pre></div> <h2 class="relative group"><a id="next-steps" class="header-link block pr-1.5 text-lg no-hover:hidden with-hover:absolute with-hover:p-1.5 with-hover:opacity-0 with-hover:group-hover:opacity-100 with-hover:right-full" href="#next-steps"><span><svg class="" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" aria-hidden="true" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 256 256"><path d="M167.594 88.393a8.001 8.001 0 0 1 0 11.314l-67.882 67.882a8 8 0 1 1-11.314-11.315l67.882-67.881a8.003 8.003 0 0 1 11.314 0zm-28.287 84.86l-28.284 28.284a40 40 0 0 1-56.567-56.567l28.284-28.284a8 8 0 0 0-11.315-11.315l-28.284 28.284a56 56 0 0 0 79.196 79.197l28.285-28.285a8 8 0 1 0-11.315-11.314zM212.852 43.14a56.002 56.002 0 0 0-79.196 0l-28.284 28.284a8 8 0 1 0 11.314 11.314l28.284-28.284a40 40 0 0 1 56.568 56.567l-28.285 28.285a8 8 0 0 0 11.315 11.314l28.284-28.284a56.065 56.065 0 0 0 0-79.196z" fill="currentColor"></path></svg></span></a> <span>Next steps</span></h2> <ul data-svelte-h="svelte-15oe4jk"><li>For more real-world examples, see how <a href="https://huggingface.co/docs/trl/main/en/gspo_token#trl.GRPOTrainer" rel="nofollow">GRPOTrainer</a> and <a href="https://huggingface.co/docs/trl/main/en/bema_for_reference_model#trl.DPOTrainer" rel="nofollow">DPOTrainer</a> extend <a href="/docs/transformers/pr_42227/en/main_classes/trainer#transformers.Trainer">Trainer</a> in TRL, or how <a href="https://github.com/axolotl-ai-cloud/axolotl/tree/main/src/axolotl/core/trainers" rel="nofollow">Axolotl</a> builds custom trainers on top of it.</li> <li>Check the <a href="./trainer_callbacks">Callbacks</a> guide if you only need to customize what happens during a training event such as logging metrics at the end of a training step.</li></ul> <a class="!text-gray-400 !no-underline text-sm flex items-center not-prose mt-4" href="https://github.com/huggingface/transformers/blob/main/docs/source/en/trainer_customize.md" target="_blank"><svg class="mr-1" xmlns="http://www.w3.org/2000/svg" aria-hidden="true" fill="currentColor" focusable="false" role="img" width="1em" height="1em" preserveAspectRatio="xMidYMid meet" viewBox="0 0 32 32"><path d="M31,16l-7,7l-1.41-1.41L28.17,16l-5.58-5.59L24,9l7,7z"></path><path d="M1,16l7-7l1.41,1.41L3.83,16l5.58,5.59L8,23l-7-7z"></path><path d="M12.419,25.484L17.639,6.552l1.932,0.518L14.351,26.002z"></path></svg> <span data-svelte-h="svelte-zjs2n5"><span class="underline">Update</span> on GitHub</span></a> <p></p>
<script>
{
__sveltekit_13nnzth = {
assets: "/docs/transformers/pr_42227/en",
base: "/docs/transformers/pr_42227/en",
env: {}
};
const element = document.currentScript.parentElement;
const data = [null,null];
Promise.all([
import("/docs/transformers/pr_42227/en/_app/immutable/entry/start.a067d14a.js"),
import("/docs/transformers/pr_42227/en/_app/immutable/entry/app.95d76f53.js")
]).then(([kit, app]) => {
kit.start(app, element, {
node_ids: [0, 672],
data,
form: null,
error: null
});
});
}
</script>

Xet Storage Details

Size:
23.1 kB
·
Xet hash:
a8e033c7a0cee4396ff35f1a0b74b731d18e391292f35b74be6b2d6b1efc159b

Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.