Hugging Face's logo

Buckets:
hf-doc-build
/
doc

Files
xet
hf-doc-build/doc / trl /v0.7.10 /en /sentiment_tuning.html
download
raw
27.1 kB
<meta charset="utf-8" /><meta name="hf:doc:metadata" content="{&quot;title&quot;:&quot;Sentiment Tuning Examples&quot;,&quot;local&quot;:&quot;sentiment-tuning-examples&quot;,&quot;sections&quot;:[{&quot;title&quot;:&quot;Usage&quot;,&quot;local&quot;:&quot;usage&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;Few notes on multi-GPU&quot;,&quot;local&quot;:&quot;few-notes-on-multi-gpu&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;Benchmarks&quot;,&quot;local&quot;:&quot;benchmarks&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;With and without gradient accumulation&quot;,&quot;local&quot;:&quot;with-and-without-gradient-accumulation&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;Comparing different models (gpt2, gpt2-xl, falcon, llama2)&quot;,&quot;local&quot;:&quot;comparing-different-models-gpt2-gpt2-xl-falcon-llama2&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;With and without PEFT&quot;,&quot;local&quot;:&quot;with-and-without-peft&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2}],&quot;depth&quot;:1}">
 <link href="/docs/trl/v0.7.10/en/_app/immutable/assets/0.e3b0c442.css" rel="modulepreload">
 <link rel="modulepreload" href="/docs/trl/v0.7.10/en/_app/immutable/entry/start.d9a24ea1.js">
 <link rel="modulepreload" href="/docs/trl/v0.7.10/en/_app/immutable/chunks/scheduler.9039eef2.js">
 <link rel="modulepreload" href="/docs/trl/v0.7.10/en/_app/immutable/chunks/singletons.9eef12cc.js">
 <link rel="modulepreload" href="/docs/trl/v0.7.10/en/_app/immutable/chunks/paths.1355483e.js">
 <link rel="modulepreload" href="/docs/trl/v0.7.10/en/_app/immutable/entry/app.5bef33b8.js">
 <link rel="modulepreload" href="/docs/trl/v0.7.10/en/_app/immutable/chunks/index.ded8f90d.js">
 <link rel="modulepreload" href="/docs/trl/v0.7.10/en/_app/immutable/nodes/0.abccdcd8.js">
 <link rel="modulepreload" href="/docs/trl/v0.7.10/en/_app/immutable/chunks/each.e59479a4.js">
 <link rel="modulepreload" href="/docs/trl/v0.7.10/en/_app/immutable/nodes/20.1afc48e0.js">
 <link rel="modulepreload" href="/docs/trl/v0.7.10/en/_app/immutable/chunks/CodeBlock.8580f3e8.js">
 <link rel="modulepreload" href="/docs/trl/v0.7.10/en/_app/immutable/chunks/Heading.f027f30d.js"><!-- HEAD_svelte-u9bgzb_START --><meta name="hf:doc:metadata" content="{&quot;title&quot;:&quot;Sentiment Tuning Examples&quot;,&quot;local&quot;:&quot;sentiment-tuning-examples&quot;,&quot;sections&quot;:[{&quot;title&quot;:&quot;Usage&quot;,&quot;local&quot;:&quot;usage&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;Few notes on multi-GPU&quot;,&quot;local&quot;:&quot;few-notes-on-multi-gpu&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;Benchmarks&quot;,&quot;local&quot;:&quot;benchmarks&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;With and without gradient accumulation&quot;,&quot;local&quot;:&quot;with-and-without-gradient-accumulation&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;Comparing different models (gpt2, gpt2-xl, falcon, llama2)&quot;,&quot;local&quot;:&quot;comparing-different-models-gpt2-gpt2-xl-falcon-llama2&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2},{&quot;title&quot;:&quot;With and without PEFT&quot;,&quot;local&quot;:&quot;with-and-without-peft&quot;,&quot;sections&quot;:[],&quot;depth&quot;:2}],&quot;depth&quot;:1}"><!-- HEAD_svelte-u9bgzb_END --> <p></p> <h1 class="relative group"><a id="sentiment-tuning-examples" 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="#sentiment-tuning-examples"><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>Sentiment Tuning Examples</span></h1> <p data-svelte-h="svelte-fb2yxu">The notebooks and scripts in this examples show how to fine-tune a model with a sentiment classifier (such as <code>lvwerra/distilbert-imdb</code>).</p> <p data-svelte-h="svelte-k5thfi">Here’s an overview of the notebooks and scripts in the <a href="https://github.com/huggingface/trl/tree/main/examples" rel="nofollow">trl repository</a>:</p> <table data-svelte-h="svelte-18q2eca"><thead><tr><th>File</th> <th>Description</th></tr></thead> <tbody><tr><td><a href="https://github.com/huggingface/trl/blob/main/examples/scripts/ppo.py" rel="nofollow"><code>examples/scripts/ppo.py</code></a> <a href="https://colab.research.google.com/github/huggingface/trl/blob/main/examples/sentiment/notebooks/gpt2-sentiment.ipynb" rel="nofollow"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"></a></td> <td>This script shows how to use the <code>PPOTrainer</code> to fine-tune a sentiment analysis model using IMDB dataset</td></tr> <tr><td><a href="https://github.com/huggingface/trl/tree/main/examples/notebooks/gpt2-sentiment.ipynb" rel="nofollow"><code>examples/notebooks/gpt2-sentiment.ipynb</code></a></td> <td>This notebook demonstrates how to reproduce the GPT2 imdb sentiment tuning example on a jupyter notebook.</td></tr> <tr><td><a href="https://github.com/huggingface/trl/tree/main/examples/notebooks/gpt2-control.ipynb" rel="nofollow"><code>examples/notebooks/gpt2-control.ipynb</code></a> <a href="https://colab.research.google.com/github/huggingface/trl/blob/main/examples/sentiment/notebooks/gpt2-sentiment-control.ipynb" rel="nofollow"><img src="https://colab.research.google.com/assets/colab-badge.svg" alt="Open In Colab"></a></td> <td>This notebook demonstrates how to reproduce the GPT2 sentiment control example on a jupyter notebook.</td></tr></tbody></table> <h2 class="relative group"><a id="usage" 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="#usage"><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>Usage</span></h2> <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-comment"># 1. run directly</span>
python examples/scripts/ppo.py
<span class="hljs-comment"># 2. run via `accelerate` (recommended), enabling more features (e.g., multiple GPUs, deepspeed)</span>
accelerate config <span class="hljs-comment"># will prompt you to define the training configuration</span>
accelerate launch examples/scripts/ppo.py <span class="hljs-comment"># launches training</span>
<span class="hljs-comment"># 3. get help text and documentation</span>
python examples/scripts/ppo.py --<span class="hljs-built_in">help</span>
<span class="hljs-comment"># 4. configure logging with wandb and, say, mini_batch_size=1 and gradient_accumulation_steps=16</span>
python examples/scripts/ppo.py --ppo_config.log_with wandb --ppo_config.mini_batch_size 1 --ppo_config.gradient_accumulation_steps 16<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-vkwr34">Note: if you don’t want to log with <code>wandb</code> remove <code>log_with=&quot;wandb&quot;</code> in the scripts/notebooks. You can also replace it with your favourite experiment tracker that’s <a href="https://huggingface.co/docs/accelerate/usage_guides/tracking" rel="nofollow">supported by <code>accelerate</code></a>.</p> <h2 class="relative group"><a id="few-notes-on-multi-gpu" 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="#few-notes-on-multi-gpu"><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>Few notes on multi-GPU</span></h2> <p data-svelte-h="svelte-qlov4x">To run in multi-GPU setup with DDP (distributed Data Parallel) change the <code>device_map</code> value to <code>device_map={&quot;&quot;: Accelerator().process_index}</code> and make sure to run your script with <code>accelerate launch yourscript.py</code>. If you want to apply naive pipeline parallelism you can use <code>device_map=&quot;auto&quot;</code>.</p> <h2 class="relative group"><a id="benchmarks" 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="#benchmarks"><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>Benchmarks</span></h2> <p data-svelte-h="svelte-di3wt5">Below are some benchmark results for <code>examples/scripts/ppo.py</code>. To reproduce locally, please check out the <code>--command</code> arguments below.</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 -->python benchmark/benchmark.py \
 --<span class="hljs-built_in">command</span> <span class="hljs-string">&quot;python examples/scripts/ppo.py --ppo_config.log_with wandb&quot;</span> \
 --num-seeds 5 \
 --start-seed 1 \
 --workers 10 \
 --slurm-nodes 1 \
 --slurm-gpus-per-task 1 \
 --slurm-ntasks 1 \
 --slurm-total-cpus 12 \
 --slurm-template-path benchmark/trl.slurm_template<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-153ri11"><img src="https://huggingface.co/datasets/trl-internal-testing/example-images/resolve/main/images/benchmark/v0.4.7-55-g110e672/sentiment.png"></p> <h2 class="relative group"><a id="with-and-without-gradient-accumulation" 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="#with-and-without-gradient-accumulation"><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>With and without gradient accumulation</span></h2> <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 -->python benchmark/benchmark.py \
 --<span class="hljs-built_in">command</span> <span class="hljs-string">&quot;python examples/scripts/ppo.py --ppo_config.exp_name sentiment_tuning_step_grad_accu --ppo_config.mini_batch_size 1 --ppo_config.gradient_accumulation_steps 128 --ppo_config.log_with wandb&quot;</span> \
 --num-seeds 5 \
 --start-seed 1 \
 --workers 10 \
 --slurm-nodes 1 \
 --slurm-gpus-per-task 1 \
 --slurm-ntasks 1 \
 --slurm-total-cpus 12 \
 --slurm-template-path benchmark/trl.slurm_template<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-x9m93d"><img src="https://huggingface.co/datasets/trl-internal-testing/example-images/resolve/main/images/benchmark/v0.4.7-55-g110e672/gradient_accu.png"></p> <h2 class="relative group"><a id="comparing-different-models-gpt2-gpt2-xl-falcon-llama2" 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="#comparing-different-models-gpt2-gpt2-xl-falcon-llama2"><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>Comparing different models (gpt2, gpt2-xl, falcon, llama2)</span></h2> <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 -->python benchmark/benchmark.py \
 --<span class="hljs-built_in">command</span> <span class="hljs-string">&quot;python examples/scripts/ppo.py --ppo_config.exp_name sentiment_tuning_gpt2 --ppo_config.log_with wandb&quot;</span> \
 --num-seeds 5 \
 --start-seed 1 \
 --workers 10 \
 --slurm-nodes 1 \
 --slurm-gpus-per-task 1 \
 --slurm-ntasks 1 \
 --slurm-total-cpus 12 \
 --slurm-template-path benchmark/trl.slurm_template
python benchmark/benchmark.py \
 --<span class="hljs-built_in">command</span> <span class="hljs-string">&quot;python examples/scripts/ppo.py --ppo_config.exp_name sentiment_tuning_gpt2xl_grad_accu --ppo_config.model_name gpt2-xl --ppo_config.mini_batch_size 16 --ppo_config.gradient_accumulation_steps 8 --ppo_config.log_with wandb&quot;</span> \
 --num-seeds 5 \
 --start-seed 1 \
 --workers 10 \
 --slurm-nodes 1 \
 --slurm-gpus-per-task 1 \
 --slurm-ntasks 1 \
 --slurm-total-cpus 12 \
 --slurm-template-path benchmark/trl.slurm_template
python benchmark/benchmark.py \
 --<span class="hljs-built_in">command</span> <span class="hljs-string">&quot;python examples/scripts/ppo.py --ppo_config.exp_name sentiment_tuning_falcon_rw_1b --ppo_config.model_name tiiuae/falcon-rw-1b --ppo_config.log_with wandb&quot;</span> \
 --num-seeds 5 \
 --start-seed 1 \
 --workers 10 \
 --slurm-nodes 1 \
 --slurm-gpus-per-task 1 \
 --slurm-ntasks 1 \
 --slurm-total-cpus 12 \
 --slurm-template-path benchmark/trl.slurm_template<!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-1g04q9s"><img src="https://huggingface.co/datasets/trl-internal-testing/example-images/resolve/main/images/benchmark/v0.4.7-55-g110e672/different_models.png"></p> <h2 class="relative group"><a id="with-and-without-peft" 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="#with-and-without-peft"><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>With and without PEFT</span></h2> <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-comment">python benchmark/benchmark</span><span class="hljs-string">.</span><span class="hljs-comment">py \</span>
 <span class="hljs-literal">--</span><span class="hljs-comment">command &quot;python examples/scripts/ppo</span><span class="hljs-string">.</span><span class="hljs-comment">py</span> <span class="hljs-literal">--</span><span class="hljs-comment">ppo_config</span><span class="hljs-string">.</span><span class="hljs-comment">exp_name sentiment_tuning_peft</span> <span class="hljs-literal">--</span><span class="hljs-comment">use_peft</span> <span class="hljs-literal">--</span><span class="hljs-comment">ppo_config</span><span class="hljs-string">.</span><span class="hljs-comment">log_with wandb&quot; \</span>
 <span class="hljs-literal">--</span><span class="hljs-comment">num</span><span class="hljs-literal">-</span><span class="hljs-comment">seeds 5 \</span>
 <span class="hljs-literal">--</span><span class="hljs-comment">start</span><span class="hljs-literal">-</span><span class="hljs-comment">seed 1 \</span>
 <span class="hljs-literal">--</span><span class="hljs-comment">workers 10 \</span>
 <span class="hljs-literal">--</span><span class="hljs-comment">slurm</span><span class="hljs-literal">-</span><span class="hljs-comment">nodes 1 \</span>
 <span class="hljs-literal">--</span><span class="hljs-comment">slurm</span><span class="hljs-literal">-</span><span class="hljs-comment">gpus</span><span class="hljs-literal">-</span><span class="hljs-comment">per</span><span class="hljs-literal">-</span><span class="hljs-comment">task 1 \</span>
 <span class="hljs-literal">--</span><span class="hljs-comment">slurm</span><span class="hljs-literal">-</span><span class="hljs-comment">ntasks 1 \</span>
 <span class="hljs-literal">--</span><span class="hljs-comment">slurm</span><span class="hljs-literal">-</span><span class="hljs-comment">total</span><span class="hljs-literal">-</span><span class="hljs-comment">cpus 12 \</span>
 <span class="hljs-literal">--</span><span class="hljs-comment">slurm</span><span class="hljs-literal">-</span><span class="hljs-comment">template</span><span class="hljs-literal">-</span><span class="hljs-comment">path benchmark/trl</span><span class="hljs-string">.</span><span class="hljs-comment">slurm_template</span><!-- HTML_TAG_END --></pre></div> <p data-svelte-h="svelte-n3ffx9"><img src="https://huggingface.co/datasets/trl-internal-testing/example-images/resolve/main/images/benchmark/v0.4.7-55-g110e672/peft.png"></p> <p></p>
<script>
 {
 __sveltekit_78hn1s = {
 assets: "/docs/trl/v0.7.10/en",
 base: "/docs/trl/v0.7.10/en",
 env: {}
 };
 const element = document.currentScript.parentElement;
 const data = [null,null];
 Promise.all([
 import("/docs/trl/v0.7.10/en/_app/immutable/entry/start.d9a24ea1.js"),
 import("/docs/trl/v0.7.10/en/_app/immutable/entry/app.5bef33b8.js")
 ]).then(([kit, app]) => {
 kit.start(app, element, {
 node_ids: [0, 20],
 data,
 form: null,
 error: null
 });
 });
 }
 </script>

Xet Storage Details

Size:
27.1 kB
·
Xet hash:
059447449ca0e2f06ccee5ee6c449201aee8ec7e3029032ef4778edf041812fb

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