Deploying to gh-pages from @ 628bca3 🚀

pytorch · Mar 20, 2024 · d78881c · d78881c
1 parent c2cf494
commit d78881c
Show file tree

Hide file tree

Showing 3 changed files with 3 additions and 23 deletions.
diff --git a/_modules/fbgemm_gpu/split_table_batched_embeddings_ops_training.html b/_modules/fbgemm_gpu/split_table_batched_embeddings_ops_training.html
@@ -728,13 +728,12 @@ <h1>Source code for fbgemm_gpu.split_table_batched_embeddings_ops_training</h1><
         <span class="n">max_gradient</span><span class="p">:</span> <span class="nb">float</span> <span class="o">=</span> <span class="mf">1.0</span><span class="p">,</span>
         <span class="n">max_norm</span><span class="p">:</span> <span class="nb">float</span> <span class="o">=</span> <span class="mf">0.0</span><span class="p">,</span>
         <span class="n">learning_rate</span><span class="p">:</span> <span class="nb">float</span> <span class="o">=</span> <span class="mf">0.01</span><span class="p">,</span>
-        <span class="c1"># used by EXACT_ADAGRAD, EXACT_ROWWISE_ADAGRAD, EXACT_ROWWISE_WEIGHTED_ADAGRAD, LAMB, and ADAM only</span>
+        <span class="c1"># used by EXACT_ADAGRAD, EXACT_ROWWISE_ADAGRAD, LAMB, and ADAM only</span>
         <span class="c1"># NOTE that default is different from nn.optim.Adagrad default of 1e-10</span>
         <span class="n">eps</span><span class="p">:</span> <span class="nb">float</span> <span class="o">=</span> <span class="mf">1.0e-8</span><span class="p">,</span>
         <span class="n">momentum</span><span class="p">:</span> <span class="nb">float</span> <span class="o">=</span> <span class="mf">0.9</span><span class="p">,</span>  <span class="c1"># used by LARS-SGD</span>
         <span class="c1"># EXACT_ADAGRAD, SGD, EXACT_SGD do not support weight decay</span>
         <span class="c1"># LAMB, ADAM, PARTIAL_ROWWISE_ADAM, PARTIAL_ROWWISE_LAMB, LARS_SGD support decoupled weight decay</span>
-        <span class="c1"># EXACT_ROWWISE_WEIGHTED_ADAGRAD supports L2 weight decay</span>
         <span class="c1"># EXACT_ROWWISE_ADAGRAD support both L2 and decoupled weight decay (via weight_decay_mode)</span>
         <span class="n">weight_decay</span><span class="p">:</span> <span class="nb">float</span> <span class="o">=</span> <span class="mf">0.0</span><span class="p">,</span>
         <span class="n">weight_decay_mode</span><span class="p">:</span> <span class="n">WeightDecayMode</span> <span class="o">=</span> <span class="n">WeightDecayMode</span><span class="o">.</span><span class="n">NONE</span><span class="p">,</span>
@@ -971,15 +970,13 @@ <h1>Source code for fbgemm_gpu.split_table_batched_embeddings_ops_training</h1><
             <span class="k">assert</span> <span class="n">optimizer</span> <span class="ow">in</span> <span class="p">(</span>
                 <span class="n">OptimType</span><span class="o">.</span><span class="n">EXACT_ADAGRAD</span><span class="p">,</span>
                 <span class="n">OptimType</span><span class="o">.</span><span class="n">EXACT_ROWWISE_ADAGRAD</span><span class="p">,</span>
-                <span class="n">OptimType</span><span class="o">.</span><span class="n">EXACT_ROWWISE_WEIGHTED_ADAGRAD</span><span class="p">,</span>
                 <span class="n">OptimType</span><span class="o">.</span><span class="n">EXACT_SGD</span><span class="p">,</span>
             <span class="p">),</span> <span class="sa">f</span><span class="s2">&quot;Optimizer </span><span class="si">{</span><span class="n">optimizer</span><span class="si">}</span><span class="s2"> is not supported in CPU mode.&quot;</span>
         <span class="k">else</span><span class="p">:</span>
             <span class="k">assert</span> <span class="n">optimizer</span> <span class="ow">in</span> <span class="p">(</span>
                 <span class="n">OptimType</span><span class="o">.</span><span class="n">ADAM</span><span class="p">,</span>
                 <span class="n">OptimType</span><span class="o">.</span><span class="n">EXACT_ADAGRAD</span><span class="p">,</span>
                 <span class="n">OptimType</span><span class="o">.</span><span class="n">EXACT_ROWWISE_ADAGRAD</span><span class="p">,</span>
-                <span class="n">OptimType</span><span class="o">.</span><span class="n">EXACT_ROWWISE_WEIGHTED_ADAGRAD</span><span class="p">,</span>
                 <span class="n">OptimType</span><span class="o">.</span><span class="n">EXACT_SGD</span><span class="p">,</span>
                 <span class="n">OptimType</span><span class="o">.</span><span class="n">LAMB</span><span class="p">,</span>
                 <span class="n">OptimType</span><span class="o">.</span><span class="n">LARS_SGD</span><span class="p">,</span>
@@ -1075,7 +1072,6 @@ <h1>Source code for fbgemm_gpu.split_table_batched_embeddings_ops_training</h1><
             <span class="k">else</span><span class="p">:</span>
                 <span class="n">rowwise</span> <span class="o">=</span> <span class="n">optimizer</span> <span class="ow">in</span> <span class="p">[</span>
                     <span class="n">OptimType</span><span class="o">.</span><span class="n">EXACT_ROWWISE_ADAGRAD</span><span class="p">,</span>
-                    <span class="n">OptimType</span><span class="o">.</span><span class="n">EXACT_ROWWISE_WEIGHTED_ADAGRAD</span><span class="p">,</span>
                 <span class="p">]</span>
                 <span class="bp">self</span><span class="o">.</span><span class="n">_apply_split</span><span class="p">(</span>
                     <span class="n">construct_split_state</span><span class="p">(</span>
@@ -1162,7 +1158,6 @@ <h1>Source code for fbgemm_gpu.split_table_batched_embeddings_ops_training</h1><
                 <span class="p">)</span>
             <span class="k">if</span> <span class="n">optimizer</span> <span class="ow">in</span> <span class="p">(</span>
                 <span class="n">OptimType</span><span class="o">.</span><span class="n">ADAM</span><span class="p">,</span>
-                <span class="n">OptimType</span><span class="o">.</span><span class="n">EXACT_ROWWISE_WEIGHTED_ADAGRAD</span><span class="p">,</span>
                 <span class="n">OptimType</span><span class="o">.</span><span class="n">LAMB</span><span class="p">,</span>
                 <span class="n">OptimType</span><span class="o">.</span><span class="n">PARTIAL_ROWWISE_ADAM</span><span class="p">,</span>
                 <span class="n">OptimType</span><span class="o">.</span><span class="n">PARTIAL_ROWWISE_LAMB</span><span class="p">,</span>
@@ -1604,18 +1599,6 @@ <h1>Source code for fbgemm_gpu.split_table_batched_embeddings_ops_training</h1><
             <span class="bp">self</span><span class="o">.</span><span class="n">iter</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">iter</span><span class="o">.</span><span class="n">cpu</span><span class="p">()</span>
         <span class="bp">self</span><span class="o">.</span><span class="n">iter</span><span class="p">[</span><span class="mi">0</span><span class="p">]</span> <span class="o">+=</span> <span class="mi">1</span>
 
-        <span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">optimizer</span> <span class="o">==</span> <span class="n">OptimType</span><span class="o">.</span><span class="n">EXACT_ROWWISE_WEIGHTED_ADAGRAD</span><span class="p">:</span>
-            <span class="k">return</span> <span class="bp">self</span><span class="o">.</span><span class="n">_report_io_size_count</span><span class="p">(</span>
-                <span class="s2">&quot;fwd_output&quot;</span><span class="p">,</span>
-                <span class="n">invokers</span><span class="o">.</span><span class="n">lookup_rowwise_weighted_adagrad</span><span class="o">.</span><span class="n">invoke</span><span class="p">(</span>
-                    <span class="n">common_args</span><span class="p">,</span>
-                    <span class="bp">self</span><span class="o">.</span><span class="n">optimizer_args</span><span class="p">,</span>
-                    <span class="n">momentum1</span><span class="p">,</span>
-                    <span class="c1"># pyre-fixme[6]: Expected `int` for 4th param but got `Union[float,</span>
-                    <span class="c1">#  int]`.</span>
-                    <span class="bp">self</span><span class="o">.</span><span class="n">iter</span><span class="o">.</span><span class="n">item</span><span class="p">(),</span>
-                <span class="p">),</span>
-            <span class="p">)</span>
         <span class="k">if</span> <span class="bp">self</span><span class="o">.</span><span class="n">optimizer</span> <span class="o">==</span> <span class="n">OptimType</span><span class="o">.</span><span class="n">ADAM</span><span class="p">:</span>
             <span class="k">return</span> <span class="bp">self</span><span class="o">.</span><span class="n">_report_io_size_count</span><span class="p">(</span>
                 <span class="s2">&quot;fwd_output&quot;</span><span class="p">,</span>
@@ -2067,7 +2050,6 @@ <h1>Source code for fbgemm_gpu.split_table_batched_embeddings_ops_training</h1><
         <span class="n">split_optimizer_states</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">split_optimizer_states</span><span class="p">()</span>
         <span class="k">if</span> <span class="p">(</span>
             <span class="bp">self</span><span class="o">.</span><span class="n">optimizer</span> <span class="o">==</span> <span class="n">OptimType</span><span class="o">.</span><span class="n">EXACT_ROWWISE_ADAGRAD</span>
-            <span class="ow">or</span> <span class="bp">self</span><span class="o">.</span><span class="n">optimizer</span> <span class="o">==</span> <span class="n">OptimType</span><span class="o">.</span><span class="n">EXACT_ROWWISE_WEIGHTED_ADAGRAD</span>
             <span class="ow">or</span> <span class="bp">self</span><span class="o">.</span><span class="n">optimizer</span> <span class="o">==</span> <span class="n">OptimType</span><span class="o">.</span><span class="n">EXACT_ADAGRAD</span>
         <span class="p">):</span>
             <span class="n">list_of_state_dict</span> <span class="o">=</span> <span class="p">[</span>
@@ -2149,7 +2131,6 @@ <h1>Source code for fbgemm_gpu.split_table_batched_embeddings_ops_training</h1><
                     <span class="n">rowwise</span><span class="o">=</span><span class="bp">self</span><span class="o">.</span><span class="n">optimizer</span>
                     <span class="ow">in</span> <span class="p">[</span>
                         <span class="n">OptimType</span><span class="o">.</span><span class="n">EXACT_ROWWISE_ADAGRAD</span><span class="p">,</span>
-                        <span class="n">OptimType</span><span class="o">.</span><span class="n">EXACT_ROWWISE_WEIGHTED_ADAGRAD</span><span class="p">,</span>
                     <span class="p">],</span>
                 <span class="p">)</span>
             <span class="p">)</span>
@@ -2635,7 +2616,6 @@ <h1>Source code for fbgemm_gpu.split_table_batched_embeddings_ops_training</h1><
 
         <span class="n">rowwise</span> <span class="o">=</span> <span class="bp">self</span><span class="o">.</span><span class="n">optimizer</span> <span class="ow">in</span> <span class="p">[</span>
             <span class="n">OptimType</span><span class="o">.</span><span class="n">EXACT_ROWWISE_ADAGRAD</span><span class="p">,</span>
-            <span class="n">OptimType</span><span class="o">.</span><span class="n">EXACT_ROWWISE_WEIGHTED_ADAGRAD</span><span class="p">,</span>
         <span class="p">]</span>
         <span class="k">if</span> <span class="n">rowwise</span><span class="p">:</span>
             <span class="n">torch</span><span class="o">.</span><span class="n">ops</span><span class="o">.</span><span class="n">fbgemm</span><span class="o">.</span><span class="n">reset_weight_momentum</span><span class="p">(</span>

diff --git a/fbgemm_gpu-python-api/table_batched_embedding_ops.html b/fbgemm_gpu-python-api/table_batched_embedding_ops.html
@@ -395,7 +395,7 @@
 <li><p><strong>weights_precision</strong> (<em>SparseType</em><em>, </em><em>optional</em>) – Data type of embedding tables (also known as weights) (<cite>SparseType.FP32</cite>, <cite>SparseType.FP16</cite>, <cite>SparseType.INT8</cite>)</p></li>
 <li><p><strong>output_dtype</strong> (<em>SparseType</em><em>, </em><em>optional</em>) – Data type of an output tensor (<cite>SparseType.FP32</cite>, <cite>SparseType.FP16</cite>, <cite>SparseType.INT8</cite>)</p></li>
 <li><p><strong>enforce_hbm</strong> (<a class="reference external" href="https://docs.python.org/3/library/functions.html#bool" title="(in Python v3.12)"><em>bool</em></a><em>, </em><em>optional</em>) – If True, place all weights/momentums in HBM when using cache</p></li>
-<li><p><strong>optimizer</strong> (<em>OptimType</em><em>, </em><em>optional</em>) – An optimizer to use for embedding table update in the backward pass. (<cite>OptimType.ADAM</cite>, <cite>OptimType.EXACT_ADAGRAD</cite>, <cite>OptimType.EXACT_ROWWISE_ADAGRAD</cite>, <cite>OptimType.EXACT_ROWWISE_WEIGHTED_ADAGRAD</cite>, <cite>OptimType.EXACT_SGD</cite>, <cite>OptimType.LAMB</cite>, <cite>OptimType.LARS_SGD</cite>, <cite>OptimType.PARTIAL_ROWWISE_ADAM</cite>, <cite>OptimType.PARTIAL_ROWWISE_LAMB</cite>, <cite>OptimType.SGD</cite>)</p></li>
+<li><p><strong>optimizer</strong> (<em>OptimType</em><em>, </em><em>optional</em>) – An optimizer to use for embedding table update in the backward pass. (<cite>OptimType.ADAM</cite>, <cite>OptimType.EXACT_ADAGRAD</cite>, <cite>OptimType.EXACT_ROWWISE_ADAGRAD</cite>, <cite>OptimType.EXACT_SGD</cite>, <cite>OptimType.LAMB</cite>, <cite>OptimType.LARS_SGD</cite>, <cite>OptimType.PARTIAL_ROWWISE_ADAM</cite>, <cite>OptimType.PARTIAL_ROWWISE_LAMB</cite>, <cite>OptimType.SGD</cite>)</p></li>
 <li><p><strong>record_cache_metrics</strong> (<em>RecordCacheMetrics</em><em>, </em><em>optional</em>) – Record number of hits, number of requests, etc if RecordCacheMetrics.record_cache_miss_counter is True and record the similar metrics table-wise if RecordCacheMetrics.record_tablewise_cache_miss is True (default is None).</p></li>
 <li><p><strong>stochastic_rounding</strong> (<a class="reference external" href="https://docs.python.org/3/library/functions.html#bool" title="(in Python v3.12)"><em>bool</em></a><em>, </em><em>optional</em>) – If True, apply stochastic rounding for weight type that is not <cite>SparseType.FP32</cite></p></li>
 <li><p><strong>gradient_clipping</strong> (<a class="reference external" href="https://docs.python.org/3/library/functions.html#bool" title="(in Python v3.12)"><em>bool</em></a><em>, </em><em>optional</em>) – If True, apply gradient clipping</p></li>

diff --git a/searchindex.js b/searchindex.js