Buckets:

hf-doc-build
/

doc-dev

Files

xet

hf-doc-build/doc-dev / trl /pr_5313 /en /unsloth_integration.html

HuggingFaceDocBuilder

30 days ago

download

raw

31.3 kB

	<meta charset="utf-8" /><meta name="hf:doc:metadata" content="{"title":"Unsloth Integration","local":"unsloth-integration","sections":[{"title":"Key Features","local":"key-features","sections":[],"depth":2},{"title":"Installation","local":"installation","sections":[{"title":"pip install","local":"pip-install","sections":[],"depth":3},{"title":"Docker Install","local":"docker-install","sections":[],"depth":3}],"depth":2},{"title":"Training","local":"training","sections":[],"depth":2},{"title":"Saving the model","local":"saving-the-model","sections":[{"title":"Saving to GGUF","local":"saving-to-gguf","sections":[],"depth":3},{"title":"Saving to vLLM","local":"saving-to-vllm","sections":[],"depth":3}],"depth":2}],"depth":1}"/>
	<link href="/docs/trl/main/en/_app/immutable/entry/start.BZkqJbXc.js" rel="modulepreload">
	<link href="/docs/trl/main/en/_app/immutable/chunks/Xx05pit2.js" rel="modulepreload">
	<link href="/docs/trl/main/en/_app/immutable/chunks/Cgtr6oey.js" rel="modulepreload">
	<link href="/docs/trl/main/en/_app/immutable/entry/app.CTLZrs8D.js" rel="modulepreload">
	<link href="/docs/trl/main/en/_app/immutable/chunks/YRUHZxfA.js" rel="modulepreload">
	<link href="/docs/trl/main/en/_app/immutable/chunks/CF2wz0qP.js" rel="modulepreload">
	<link href="/docs/trl/main/en/_app/immutable/chunks/DsnmJJEf.js" rel="modulepreload">
	<link href="/docs/trl/main/en/_app/immutable/chunks/B4jKbUu6.js" rel="modulepreload">
	<link href="/docs/trl/main/en/_app/immutable/nodes/0.CD6H8d2I.js" rel="modulepreload">
	<link href="/docs/trl/main/en/_app/immutable/chunks/B8yBvtRr.js" rel="modulepreload">
	<link href="/docs/trl/main/en/_app/immutable/nodes/42.CdIkHQ0T.js" rel="modulepreload">
	<link href="/docs/trl/main/en/_app/immutable/chunks/DMzrT19u.js" rel="modulepreload">
	<!--y6mbuf--><meta name="hf:doc:metadata" content="{"title":"Unsloth Integration","local":"unsloth-integration","sections":[{"title":"Key Features","local":"key-features","sections":[],"depth":2},{"title":"Installation","local":"installation","sections":[{"title":"pip install","local":"pip-install","sections":[],"depth":3},{"title":"Docker Install","local":"docker-install","sections":[],"depth":3}],"depth":2},{"title":"Training","local":"training","sections":[],"depth":2},{"title":"Saving the model","local":"saving-the-model","sections":[{"title":"Saving to GGUF","local":"saving-to-gguf","sections":[],"depth":3},{"title":"Saving to vLLM","local":"saving-to-vllm","sections":[],"depth":3}],"depth":2}],"depth":1}"/><!---->
	<link href="/docs/trl/main/en/_app/immutable/assets/0.tn0RQdqM.css" rel="modulepreload"> <!--[--><!--[0--><!--[--><!--[0--><!--[--><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> <!--[-1--><!--]--></div><!----> <!--[0--><h1 class="relative group"><a id="unsloth-integration" 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="#unsloth-integration"><span><svg 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>Unsloth Integration</span></h1><!--]--><!----> <p>Unsloth is an open‑source framework for fine‑tuning and reinforcement learning that trains LLMs (like Llama, OpenAI gpt-oss, Mistral, Gemma, DeepSeek, and more) up to 2× faster with up to 80% less VRAM. Unsloth allows <a href="https://huggingface.co/docs/trl/en/unsloth_integration#Training" rel="nofollow">training</a>, evaluation, running and <a href="https://huggingface.co/docs/trl/en/unsloth_integration#Saving-the-model" rel="nofollow">deployment</a> with other inference engines like llama.cpp, Ollama and vLLM.</p> <p>The library provides a streamlined, Hugging Face compatible workflow for training, evaluation, inference and deployment and is fully compatible with <a href="/docs/trl/pr_5313/en/sft_trainer#trl.SFTTrainer">SFTTrainer</a>.</p> <!--[1--><h2 class="relative group"><a id="key-features" 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="#key-features"><span><svg 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>Key Features</span></h2><!--]--><!----> <ul><li>Training support for all transformer compatible models: Text-to-speech (TTS), multimodal, BERT, RL and more</li> <li>Supports full fine-tuning, pretraining, LoRA, QLoRA, 8-bit training & more</li> <li>Works on Linux, Windows, Colab, Kaggle; NVIDIA GPUs, soon AMD & Intel setups</li> <li>Supports most features TRL supports, including RLHF (GSPO, GRPO, DPO etc.)</li> <li>Hand-written Triton kernels and a manual backprop engine ensure no accuracy degradation (0% approximation error)</li></ul> <!--[1--><h2 class="relative group"><a id="installation" 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="#installation"><span><svg 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>Installation</span></h2><!--]--><!----> <!--[2--><h3 class="relative group"><a id="pip-install" 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="#pip-install"><span><svg 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>pip install</span></h3><!--]--><!----> <p>Local Installation (Linux recommended):</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 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="language-sh "><!---->pip install unsloth<!----></pre></div><!----> <p>You can also install <code>unsloth</code> according to the <a href="https://docs.unsloth.ai/get-started/installing-+-updating" rel="nofollow">official documentation</a>. Once installed, you can incorporate unsloth into your workflow in a very simple manner; instead of loading <a href="https://huggingface.co/docs/transformers/main/en/model_doc/auto#transformers.AutoModelForCausalLM" rel="nofollow">AutoModelForCausalLM</a>, you just need to load a <code>FastLanguageModel</code> as follows:</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 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="language-python "><!----><span class="hljs-keyword">import</span> torch
	<span class="hljs-keyword">from</span> trl <span class="hljs-keyword">import</span> SFTConfig, SFTTrainer
	<span class="hljs-keyword">from</span> unsloth <span class="hljs-keyword">import</span> FastLanguageModel

	max_length = <span class="hljs-number">2048</span> <span class="hljs-comment"># Supports automatic RoPE Scaling, so choose any number</span>

	<span class="hljs-comment"># Load model</span>
	model, tokenizer = FastLanguageModel.from_pretrained(
	model_name=<span class="hljs-string">"unsloth/mistral-7b"</span>,
	max_seq_length=max_length,
	dtype=<span class="hljs-string">"auto"</span>, <span class="hljs-comment"># For auto-detection. Float16 for Tesla T4, V100, Bfloat16 for Ampere+</span>
	load_in_4bit=<span class="hljs-literal">True</span>, <span class="hljs-comment"># Use 4bit quantization to reduce memory usage. Can be False</span>
	)

	<span class="hljs-comment"># Do model patching and add fast LoRA weights</span>
	model = FastLanguageModel.get_peft_model(
	model,
	r=<span class="hljs-number">16</span>,
	target_modules=[
	<span class="hljs-string">"q_proj"</span>,
	<span class="hljs-string">"k_proj"</span>,
	<span class="hljs-string">"v_proj"</span>,
	<span class="hljs-string">"o_proj"</span>,
	<span class="hljs-string">"gate_proj"</span>,
	<span class="hljs-string">"up_proj"</span>,
	<span class="hljs-string">"down_proj"</span>,
	],
	lora_alpha=<span class="hljs-number">16</span>,
	lora_dropout=<span class="hljs-number">0</span>, <span class="hljs-comment"># Dropout = 0 is currently optimized</span>
	bias=<span class="hljs-string">"none"</span>, <span class="hljs-comment"># Bias = "none" is currently optimized</span>
	use_gradient_checkpointing=<span class="hljs-literal">True</span>,
	random_state=<span class="hljs-number">3407</span>,
	)

	training_args = SFTConfig(output_dir=<span class="hljs-string">"./output"</span>, max_length=max_length)

	trainer = SFTTrainer(
	model=model,
	args=training_args,
	train_dataset=dataset,
	)
	trainer.train()<!----></pre></div><!----> <p>The saved model is fully compatible with Hugging Face’s transformers library. Learn more about unsloth in their <a href="https://github.com/unslothai/unsloth" rel="nofollow">official repository</a>.</p> <!--[2--><h3 class="relative group"><a id="docker-install" 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="#docker-install"><span><svg 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>Docker Install</span></h3><!--]--><!----> <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 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="language-sh "><!---->docker run -d -e JUPYTER_PASSWORD=<span class="hljs-string">"mypassword"</span> \
	-p 8888:8888 -p 2222:22 \
	-v $(<span class="hljs-built_in">pwd</span>)/work:/workspace/work \
	--gpus all \
	unsloth/unsloth<!----></pre></div><!----> <p>Access Jupyter Lab at <code>http://localhost:8888</code> and start fine-tuning!</p> <!--[1--><h2 class="relative group"><a id="training" 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="#training"><span><svg 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>Training</span></h2><!--]--><!----> <p>These are some core settings you can toggle before training:</p> <ul><li><code>max_seq_length = 2048</code> – Controls context length. While Llama-3 supports 8192, we recommend 2048 for testing. Unsloth enables 4× longer context fine-tuning.</li> <li><code>dtype = "auto"</code> – For auto-detection; use torch.float16 or torch.bfloat16 for newer GPUs.</li> <li><code>load_in_4bit = True</code> – Enables 4-bit quantization, reducing memory use 4× for fine-tuning. Disabling it allows for LoRA 16-bit fine-tuning to be enabled.</li> <li>To enable full fine-tuning (FFT), set <code>full_finetuning = True</code>. For 8-bit fine-tuning, set <code>load_in_8bit = True</code>. Note: Only one training method can be set to True at a time.</li></ul> <p>For more information on configuring Unsloth’s hyperparameters and features, read their <a href="https://docs.unsloth.ai/get-started/fine-tuning-llms-guide" rel="nofollow">documentation guide here</a>.</p> <!--[1--><h2 class="relative group"><a id="saving-the-model" 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="#saving-the-model"><span><svg 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>Saving the model</span></h2><!--]--><!----> <p>Unsloth allows you to directly save the finetuned model as a small file called a LoRA adapter. You can instead push to the Hugging Face hub as well if you want to upload your model! Remember to get a <a href="https://huggingface.co/settings/tokens" rel="nofollow">Hugging Face token</a> and add your token!</p> <!--[2--><h3 class="relative group"><a id="saving-to-gguf" 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="#saving-to-gguf"><span><svg 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>Saving to GGUF</span></h3><!--]--><!----> <p>To save to GGUF, Unsloth uses llama.cpp. To save locally:</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 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="language-python "><!---->model.save_pretrained_gguf(<span class="hljs-string">"directory"</span>, tokenizer, quantization_method = <span class="hljs-string">"q4_k_m"</span>)
	model.save_pretrained_gguf(<span class="hljs-string">"directory"</span>, tokenizer, quantization_method = <span class="hljs-string">"q8_0"</span>)
	model.save_pretrained_gguf(<span class="hljs-string">"directory"</span>, tokenizer, quantization_method = <span class="hljs-string">"f16"</span>)<!----></pre></div><!----> <p>To push to the hub:</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 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="language-python "><!---->model.push_to_hub_gguf(<span class="hljs-string">"hf_username/directory"</span>, tokenizer, quantization_method = <span class="hljs-string">"q4_k_m"</span>)
	model.push_to_hub_gguf(<span class="hljs-string">"hf_username/directory"</span>, tokenizer, quantization_method = <span class="hljs-string">"q8_0"</span>)<!----></pre></div><!----> <!--[2--><h3 class="relative group"><a id="saving-to-vllm" 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="#saving-to-vllm"><span><svg 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>Saving to vLLM</span></h3><!--]--><!----> <p>To save to 16-bit for vLLM, use:</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 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="language-python "><!---->model.save_pretrained_merged(<span class="hljs-string">"model"</span>, tokenizer, save_method = <span class="hljs-string">"merged_16bit"</span>,)
	model.push_to_hub_merged(<span class="hljs-string">"hf/model"</span>, tokenizer, save_method = <span class="hljs-string">"merged_16bit"</span>, token = <span class="hljs-string">""</span>)<!----></pre></div><!----> <a class="!text-gray-400 !no-underline text-sm flex items-center not-prose mt-4" href="https://github.com/huggingface/trl/blob/main/docs/source/unsloth_integration.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><span class="underline">Update</span> on GitHub</span></a><!----> <p></p><!--]--><!----><!--]--><!--]--><!--]--> <!--[-1--><!--]--><!--]-->

	<script>
	{
	__sveltekit_1to0zqp = {
	base: "/docs/trl/main/en",
	assets: "/docs/trl/main/en"
	};

	const element = document.currentScript.parentElement;

	Promise.all([
	import("/docs/trl/main/en/_app/immutable/entry/start.BZkqJbXc.js"),
	import("/docs/trl/main/en/_app/immutable/entry/app.CTLZrs8D.js")
	]).then(([kit, app]) => {
	kit.start(app, element, {
	node_ids: [0, 42],
	data: [null,null],
	form: null,
	error: null
	});
	});
	}
	</script>

Xet Storage Details

Size:: 31.3 kB
Xet hash:: 304267ab33b70fe41243730c99807ee583d6f414997e1ef5367b975be5aa23aa

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