auto-generating sphinx docs

Author: pytorchbot

main/_modules/torchtune/modules/peft/_utils.html

@@ -457,7 +457,6 @@ <h1>Source code for torchtune.modules.peft._utils</h1><div class="highlight"><pr
 
 <span class="kn">import</span><span class="w"> </span><span class="nn">torch</span>
 <span class="kn">from</span><span class="w"> </span><span class="nn">torch</span><span class="w"> </span><span class="kn">import</span> <span class="n">nn</span>
-<span class="kn">from</span><span class="w"> </span><span class="nn">torchtune.utils._logging</span><span class="w"> </span><span class="kn">import</span> <span class="n">deprecated</span>
 
 <span class="c1"># Modules from MultiHeadAttention that LoRA can be applied to</span>
 <span class="n">LORA_ATTN_MODULES</span> <span class="o">=</span> <span class="n">Literal</span><span class="p">[</span><span class="s2">&quot;q_proj&quot;</span><span class="p">,</span> <span class="s2">&quot;k_proj&quot;</span><span class="p">,</span> <span class="s2">&quot;v_proj&quot;</span><span class="p">,</span> <span class="s2">&quot;output_proj&quot;</span><span class="p">]</span>
@@ -761,20 +760,6 @@ <h1>Source code for torchtune.modules.peft._utils</h1><div class="highlight"><pr
                 <span class="k">raise</span> <span class="ne">AssertionError</span><span class="p">(</span><span class="sa">f</span><span class="s2">&quot;Missing LoRA key </span><span class="si">{</span><span class="n">k</span><span class="si">}</span><span class="s2"> from adapter state dict&quot;</span><span class="p">)</span>
     <span class="k">if</span> <span class="n">lora_unexpected</span><span class="p">:</span>
         <span class="k">raise</span> <span class="ne">AssertionError</span><span class="p">(</span><span class="s2">&quot;Unexpected key loading adapter&quot;</span><span class="p">)</span></div>
-
-
-<span class="nd">@deprecated</span><span class="p">(</span>
-    <span class="n">msg</span><span class="o">=</span><span class="s2">&quot;load_dora_magnitudes will be deprecated in 0.6.0. Please use DoRALinear.initialize_dora_magnitude instead.&quot;</span>
-<span class="p">)</span>
-<span class="k">def</span><span class="w"> </span><span class="nf">load_dora_magnitudes</span><span class="p">(</span><span class="n">model</span><span class="p">:</span> <span class="n">nn</span><span class="o">.</span><span class="n">Module</span><span class="p">)</span> <span class="o">-&gt;</span> <span class="kc">None</span><span class="p">:</span>
-<span class="w">    </span><span class="sd">&quot;&quot;&quot;</span>
-<span class="sd">    For DoRA magnitude we use setattr to move from meta device</span>
-<span class="sd">    &quot;&quot;&quot;</span>
-    <span class="n">dora_parents</span> <span class="o">=</span> <span class="p">{</span>
-        <span class="n">n</span><span class="p">:</span> <span class="n">p</span> <span class="k">for</span> <span class="n">n</span><span class="p">,</span> <span class="n">p</span> <span class="ow">in</span> <span class="n">model</span><span class="o">.</span><span class="n">named_modules</span><span class="p">()</span> <span class="k">if</span> <span class="nb">hasattr</span><span class="p">(</span><span class="n">p</span><span class="p">,</span> <span class="s2">&quot;adapter_params&quot;</span><span class="p">)</span>
-    <span class="p">}</span>
-    <span class="n">sd</span> <span class="o">=</span> <span class="p">{</span><span class="sa">f</span><span class="s2">&quot;</span><span class="si">{</span><span class="n">n</span><span class="si">}</span><span class="s2">.magnitude&quot;</span><span class="p">:</span> <span class="n">p</span><span class="o">.</span><span class="n">magnitude</span> <span class="k">for</span> <span class="n">n</span><span class="p">,</span> <span class="n">p</span> <span class="ow">in</span> <span class="n">dora_parents</span><span class="o">.</span><span class="n">items</span><span class="p">()}</span>
-    <span class="n">model</span><span class="o">.</span><span class="n">load_state_dict</span><span class="p">(</span><span class="n">sd</span><span class="p">,</span> <span class="n">strict</span><span class="o">=</span><span class="kc">False</span><span class="p">,</span> <span class="n">assign</span><span class="o">=</span><span class="kc">True</span><span class="p">)</span>
 </pre></div>
 
              </article>

main/_modules/torchtune/training/quantization.html

@@ -453,7 +453,6 @@ <h1>Source code for torchtune.training.quantization</h1><div class="highlight"><
 <span class="c1"># LICENSE file in the root directory of this source tree.</span>
 
 <span class="kn">from</span><span class="w"> </span><span class="nn">typing</span><span class="w"> </span><span class="kn">import</span> <span class="n">Callable</span><span class="p">,</span> <span class="n">Optional</span>
-<span class="kn">from</span><span class="w"> </span><span class="nn">warnings</span><span class="w"> </span><span class="kn">import</span> <span class="n">warn</span>
 
 <span class="kn">from</span><span class="w"> </span><span class="nn">torch</span><span class="w"> </span><span class="kn">import</span> <span class="n">nn</span>
 <span class="kn">from</span><span class="w"> </span><span class="nn">torchtune.modules.peft.lora</span><span class="w"> </span><span class="kn">import</span> <span class="n">LoRALinear</span><span class="p">,</span> <span class="n">QATLoRALinear</span>
@@ -592,6 +591,7 @@ <h1>Source code for torchtune.training.quantization</h1><div class="highlight"><
     <span class="s2">&quot;4w-qat-module-swap&quot;</span>
 <span class="p">]</span> <span class="o">=</span> <span class="n">enable_4w_fake_quant_module_swap</span>
 
+
 <span class="c1"># int8 dynamic activations + int4 weight</span>
 <span class="k">class</span><span class="w"> </span><span class="nc">Int8DynActInt4WeightQATQuantizerModuleSwap</span><span class="p">(</span><span class="n">Int8DynActInt4WeightQATQuantizer</span><span class="p">):</span>
     <span class="k">pass</span>
@@ -627,12 +627,7 @@ <h1>Source code for torchtune.training.quantization</h1><div class="highlight"><
 <span class="sd">    Returns:</span>
 <span class="sd">        Optional[str]: The quantization mode.</span>
 <span class="sd">    &quot;&quot;&quot;</span>
-    <span class="n">mode</span> <span class="o">=</span> <span class="n">_quantizer_to_mode</span><span class="o">.</span><span class="n">get</span><span class="p">(</span><span class="nb">type</span><span class="p">(</span><span class="n">quantizer</span><span class="p">),</span> <span class="kc">None</span><span class="p">)</span>
-    <span class="k">if</span> <span class="n">mode</span> <span class="ow">is</span> <span class="ow">not</span> <span class="kc">None</span> <span class="ow">and</span> <span class="s2">&quot;module-swap&quot;</span> <span class="ow">in</span> <span class="n">mode</span><span class="p">:</span>
-        <span class="n">warn</span><span class="p">(</span>
-            <span class="s2">&quot;*QuantizerModuleSwap is deprecated. Please use the version without &#39;ModuleSwap&#39; instead&quot;</span>
-        <span class="p">)</span>
-    <span class="k">return</span> <span class="n">mode</span></div>
+    <span class="k">return</span> <span class="n">_quantizer_to_mode</span><span class="o">.</span><span class="n">get</span><span class="p">(</span><span class="nb">type</span><span class="p">(</span><span class="n">quantizer</span><span class="p">),</span> <span class="kc">None</span><span class="p">)</span></div>
 
 
 <span class="k">def</span><span class="w"> </span><span class="nf">_get_disable_fake_quant</span><span class="p">(</span><span class="n">quantizer_mode</span><span class="p">:</span> <span class="nb">str</span><span class="p">)</span> <span class="o">-&gt;</span> <span class="n">Callable</span><span class="p">:</span>

main/_sources/tutorials/e2e_flow.rst.txt

@@ -275,18 +275,20 @@ Let's first copy over the config to our local working directory so we can make c
 
     $ tune cp generation ./custom_generation_config.yaml
     Copied file to custom_generation_config.yaml
+    $ mkdir /tmp/torchtune/llama3_2_3B/lora_single_device/out
 
 Let's modify ``custom_generation_config.yaml`` to include the following changes. Again, you only need
  to replace two fields: ``output_dir`` and ``checkpoint_files``
 
 .. code-block:: yaml
 
-    output_dir: /tmp/torchtune/llama3_2_3B/lora_single_device/epoch_0
+    checkpoint_dir: /tmp/torchtune/llama3_2_3B/lora_single_device/epoch_0
+    output_dir: /tmp/torchtune/llama3_2_3B/lora_single_device/out
 
     # Tokenizer
     tokenizer:
         _component_: torchtune.models.llama3.llama3_tokenizer
-        path: ${output_dir}/original/tokenizer.model
+        path: ${checkpoint_dir}/original/tokenizer.model
         prompt_template: null
 
     model:
@@ -295,7 +297,7 @@ Let's modify ``custom_generation_config.yaml`` to include the following changes.
 
     checkpointer:
         _component_: torchtune.training.FullModelHFCheckpointer
-        checkpoint_dir: ${output_dir}
+        checkpoint_dir: ${checkpoint_dir}
         checkpoint_files: [
             ft-model-00001-of-00002.safetensors,
             ft-model-00002-of-00002.safetensors,
@@ -312,8 +314,8 @@ Let's modify ``custom_generation_config.yaml`` to include the following changes.
 
     # Generation arguments; defaults taken from gpt-fast
     prompt:
-    system: null
-    user: "Tell me a joke. "
+      system: null
+      user: "Tell me a joke. "
     max_new_tokens: 300
     temperature: 0.6 # 0.8 and 0.6 are popular values to try
     top_k: 300
@@ -330,7 +332,7 @@ these parameters.
 
 .. code-block:: text
 
-    $ tune run generate --config ./custom_generation_config.yaml prompt="tell me a joke. "
+    $ tune run generate --config ./custom_generation_config.yaml prompt.user="Tell me a joke. "
     Tell me a joke. Here's a joke for you:
 
     What do you call a fake noodle?

main/searchindex.js

@@ -707,18 +707,20 @@ <h3>Generate some output<a class="headerlink" href="#generate-some-output" title
 <p>Let’s first copy over the config to our local working directory so we can make changes.</p>
 <div class="highlight-text notranslate"><div class="highlight"><pre><span></span>$ tune cp generation ./custom_generation_config.yaml
 Copied file to custom_generation_config.yaml
+$ mkdir /tmp/torchtune/llama3_2_3B/lora_single_device/out
 </pre></div>
 </div>
 <dl class="simple">
 <dt>Let’s modify <code class="docutils literal notranslate"><span class="pre">custom_generation_config.yaml</span></code> to include the following changes. Again, you only need</dt><dd><p>to replace two fields: <code class="docutils literal notranslate"><span class="pre">output_dir</span></code> and <code class="docutils literal notranslate"><span class="pre">checkpoint_files</span></code></p>
 </dd>
 </dl>
-<div class="highlight-yaml notranslate"><div class="highlight"><pre><span></span><span class="nt">output_dir</span><span class="p">:</span><span class="w"> </span><span class="l l-Scalar l-Scalar-Plain">/tmp/torchtune/llama3_2_3B/lora_single_device/epoch_0</span>
+<div class="highlight-yaml notranslate"><div class="highlight"><pre><span></span><span class="nt">checkpoint_dir</span><span class="p">:</span><span class="w"> </span><span class="l l-Scalar l-Scalar-Plain">/tmp/torchtune/llama3_2_3B/lora_single_device/epoch_0</span>
+<span class="nt">output_dir</span><span class="p">:</span><span class="w"> </span><span class="l l-Scalar l-Scalar-Plain">/tmp/torchtune/llama3_2_3B/lora_single_device/out</span>
 
 <span class="c1"># Tokenizer</span>
 <span class="nt">tokenizer</span><span class="p">:</span>
 <span class="w">    </span><span class="nt">_component_</span><span class="p">:</span><span class="w"> </span><span class="l l-Scalar l-Scalar-Plain">torchtune.models.llama3.llama3_tokenizer</span>
-<span class="w">    </span><span class="nt">path</span><span class="p">:</span><span class="w"> </span><span class="l l-Scalar l-Scalar-Plain">${output_dir}/original/tokenizer.model</span>
+<span class="w">    </span><span class="nt">path</span><span class="p">:</span><span class="w"> </span><span class="l l-Scalar l-Scalar-Plain">${checkpoint_dir}/original/tokenizer.model</span>
 <span class="w">    </span><span class="nt">prompt_template</span><span class="p">:</span><span class="w"> </span><span class="l l-Scalar l-Scalar-Plain">null</span>
 
 <span class="nt">model</span><span class="p">:</span>
@@ -727,7 +729,7 @@ <h3>Generate some output<a class="headerlink" href="#generate-some-output" title
 
 <span class="nt">checkpointer</span><span class="p">:</span>
 <span class="w">    </span><span class="nt">_component_</span><span class="p">:</span><span class="w"> </span><span class="l l-Scalar l-Scalar-Plain">torchtune.training.FullModelHFCheckpointer</span>
-<span class="w">    </span><span class="nt">checkpoint_dir</span><span class="p">:</span><span class="w"> </span><span class="l l-Scalar l-Scalar-Plain">${output_dir}</span>
+<span class="w">    </span><span class="nt">checkpoint_dir</span><span class="p">:</span><span class="w"> </span><span class="l l-Scalar l-Scalar-Plain">${checkpoint_dir}</span>
 <span class="w">    </span><span class="nt">checkpoint_files</span><span class="p">:</span><span class="w"> </span><span class="p p-Indicator">[</span>
 <span class="w">        </span><span class="nv">ft-model-00001-of-00002.safetensors</span><span class="p p-Indicator">,</span>
 <span class="w">        </span><span class="nv">ft-model-00002-of-00002.safetensors</span><span class="p p-Indicator">,</span>
@@ -744,8 +746,8 @@ <h3>Generate some output<a class="headerlink" href="#generate-some-output" title
 
 <span class="c1"># Generation arguments; defaults taken from gpt-fast</span>
 <span class="nt">prompt</span><span class="p">:</span>
-<span class="nt">system</span><span class="p">:</span><span class="w"> </span><span class="l l-Scalar l-Scalar-Plain">null</span>
-<span class="nt">user</span><span class="p">:</span><span class="w"> </span><span class="s">&quot;Tell</span><span class="nv"> </span><span class="s">me</span><span class="nv"> </span><span class="s">a</span><span class="nv"> </span><span class="s">joke.</span><span class="nv"> </span><span class="s">&quot;</span>
+<span class="w">  </span><span class="nt">system</span><span class="p">:</span><span class="w"> </span><span class="l l-Scalar l-Scalar-Plain">null</span>
+<span class="w">  </span><span class="nt">user</span><span class="p">:</span><span class="w"> </span><span class="s">&quot;Tell</span><span class="nv"> </span><span class="s">me</span><span class="nv"> </span><span class="s">a</span><span class="nv"> </span><span class="s">joke.</span><span class="nv"> </span><span class="s">&quot;</span>
 <span class="nt">max_new_tokens</span><span class="p">:</span><span class="w"> </span><span class="l l-Scalar l-Scalar-Plain">300</span>
 <span class="nt">temperature</span><span class="p">:</span><span class="w"> </span><span class="l l-Scalar l-Scalar-Plain">0.6</span><span class="w"> </span><span class="c1"># 0.8 and 0.6 are popular values to try</span>
 <span class="nt">top_k</span><span class="p">:</span><span class="w"> </span><span class="l l-Scalar l-Scalar-Plain">300</span>
@@ -760,7 +762,7 @@ <h3>Generate some output<a class="headerlink" href="#generate-some-output" title
 <code class="docutils literal notranslate"><span class="pre">temperature=0.8</span></code>. These parameters control how the probabilities for
 sampling are computed. We recommend inspecting the model with these before playing around with
 these parameters.</p>
-<div class="highlight-text notranslate"><div class="highlight"><pre><span></span>$ tune run generate --config ./custom_generation_config.yaml prompt=&quot;tell me a joke. &quot;
+<div class="highlight-text notranslate"><div class="highlight"><pre><span></span>$ tune run generate --config ./custom_generation_config.yaml prompt.user=&quot;Tell me a joke. &quot;
 Tell me a joke. Here&#39;s a joke for you:
 
 What do you call a fake noodle?