spellcheck

pianpwk · pianpwk · commit a0b9760204db · 2025-01-22T16:15:55.000-08:00
diff --git a/en-wordlist.txt b/en-wordlist.txt
@@ -82,6 +82,7 @@ FX's
 FairSeq
 Fastpath
 FakeTensor
+FakeTensors
 FFN
 FloydHub
 FloydHub's
@@ -240,6 +241,7 @@ Sohn
 Spacy
 SwiGLU
 SymInt
+SymInts
 TCP
 THP
 TIAToolbox
@@ -371,6 +373,7 @@ downsamples
 dropdown
 dtensor
 dtype
+dtypes
 duration
 elementwise
 embeddings
@@ -655,7 +658,6 @@ RecSys
 TorchRec
 sharding
 TBE
-dtype
 EBC
 sharder
 hyperoptimized
diff --git a/intermediate_source/torch_export_tutorial.py b/intermediate_source/torch_export_tutorial.py
@@ -639,7 +639,7 @@ def forward(self, x, y):
 # with parts of user code where compilation relies on data values. In short, if the compiler requires a concrete, data-dependent value in order to proceed, it will error out, complaining that
 # FakeTensor tracing isn't providing the information required.
 #
-# Data-depdenent values appear in many places, and common sources are calls like ``item()``, ``tolist()``, or ``torch.unbind()`` that extract scalar values from tensors.
+# Data-dependent values appear in many places, and common sources are calls like ``item()``, ``tolist()``, or ``torch.unbind()`` that extract scalar values from tensors.
 # How are these values represented in the exported program? In the `Constraints/Dynamic Shapes <https://pytorch.org/tutorials/intermediate/torch_export_tutorial.html#constraints-dynamic-shapes>`_
 # section, we talked about allocating symbols to represent dynamic input dimensions.
 # The same happens here: we allocate symbols for every data-dependent value that appears in the program. The important distinction is that these are "unbacked" symbols or "unbacked SymInts",
@@ -687,7 +687,7 @@ def forward(self, x, y):
 
 ######################################################################
 # Here we actually need the "hint", or the concrete value of ``a`` for the compiler to decide whether to trace ``return y + 2`` or ``return y * 5`` as the output.
-# Because we trace with FakeTensors, we don't know what ``a // 2 >= 5`` actually evaluates to, and export errors out with "Could not guard on data-dependent expression ``u0 // 2 >= 5`` (unhinted)".
+# Because we trace with FakeTensors, we don't know what ``a // 2 >= 5`` actually evaluates to, and export errors out with "Could not guard on data-dependent expression ``u0 // 2 >= 5 (unhinted)``".
 #
 # So how do we actually export this? Unlike ``torch.compile()``, export requires full graph compilation, and we can't just graph break on this. Here's some basic options:
 #

Original file line number	Diff line number	Diff line change
`@@ -639,7 +639,7 @@ def forward(self, x, y):`
`639`	`639`	`# with parts of user code where compilation relies on data values. In short, if the compiler requires a concrete, data-dependent value in order to proceed, it will error out, complaining that`
`640`	`640`	`# FakeTensor tracing isn't providing the information required.`
`641`	`641`	`#`
`642`		-# Data-depdenent values appear in many places, and common sources are calls like ``item()``, ``tolist()``, or ``torch.unbind()`` that extract scalar values from tensors.
	`642`	+# Data-dependent values appear in many places, and common sources are calls like ``item()``, ``tolist()``, or ``torch.unbind()`` that extract scalar values from tensors.
`643`	`643`	# How are these values represented in the exported program? In the `Constraints/Dynamic Shapes <https://pytorch.org/tutorials/intermediate/torch_export_tutorial.html#constraints-dynamic-shapes>`_
`644`	`644`	`# section, we talked about allocating symbols to represent dynamic input dimensions.`
`645`	`645`	`# The same happens here: we allocate symbols for every data-dependent value that appears in the program. The important distinction is that these are "unbacked" symbols or "unbacked SymInts",`
`@@ -687,7 +687,7 @@ def forward(self, x, y):`
`687`	`687`
`688`	`688`	`######################################################################`
`689`	`689`	# Here we actually need the "hint", or the concrete value of ``a`` for the compiler to decide whether to trace ``return y + 2`` or ``return y * 5`` as the output.
`690`		-# Because we trace with FakeTensors, we don't know what ``a // 2 >= 5`` actually evaluates to, and export errors out with "Could not guard on data-dependent expression ``u0 // 2 >= 5`` (unhinted)".
	`690`	+# Because we trace with FakeTensors, we don't know what ``a // 2 >= 5`` actually evaluates to, and export errors out with "Could not guard on data-dependent expression ``u0 // 2 >= 5 (unhinted)``".
`691`	`691`	`#`
`692`	`692`	# So how do we actually export this? Unlike ``torch.compile()``, export requires full graph compilation, and we can't just graph break on this. Here's some basic options:
`693`	`693`	`#`