You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@tvm.apache.org by tk...@apache.org on 2023/05/19 17:48:28 UTC
[tvm] 05/08: fix tuning, comparisons
This is an automated email from the ASF dual-hosted git repository.
tkonolige pushed a commit to branch tkonolige/relax_pad_etc_new
in repository https://gitbox.apache.org/repos/asf/tvm.git
commit 9896c6a39464fcefadfa9b7e99801de89b830db7
Author: Tristan Konolige <tk...@octoml.ai>
AuthorDate: Mon May 15 23:14:36 2023 +0000
fix tuning, comparisons
---
python/tvm/relax/frontend/torch/fx_translator.py | 19 +++++++++++++++++--
python/tvm/relax/utils.py | 2 +-
src/relax/ir/emit_te.cc | 2 +-
src/tir/analysis/estimate_flops.cc | 15 ++++++++++++---
4 files changed, 31 insertions(+), 7 deletions(-)
diff --git a/python/tvm/relax/frontend/torch/fx_translator.py b/python/tvm/relax/frontend/torch/fx_translator.py
index bdf2c3375f..6222b11432 100644
--- a/python/tvm/relax/frontend/torch/fx_translator.py
+++ b/python/tvm/relax/frontend/torch/fx_translator.py
@@ -376,6 +376,18 @@ class TorchFXImporter:
lhs, rhs = self.retrieve_args(node)
return self._call_binary_op(relax.op.less, lhs, rhs)
+ def _le(self, node: fx.node.Node) -> relax.Expr:
+ lhs, rhs = self.retrieve_args(node)
+ return self._call_binary_op(relax.op.less_equal, lhs, rhs)
+
+ def _gt(self, node: fx.node.Node) -> relax.Expr:
+ lhs, rhs = self.retrieve_args(node)
+ return self._call_binary_op(relax.op.greater, lhs, rhs)
+
+ def _ge(self, node: fx.node.Node) -> relax.Expr:
+ lhs, rhs = self.retrieve_args(node)
+ return self._call_binary_op(relax.op.greater_equal, lhs, rhs)
+
def _eq(self, node: fx.node.Node) -> relax.Expr:
lhs, rhs = self.retrieve_args(node)
return self._call_binary_op(relax.op.equal, lhs, rhs)
@@ -1154,8 +1166,8 @@ class TorchFXImporter:
from torch import fx
if any([isinstance(s, fx.node.Node) for s in size]):
- size = relax.op.concat([self.env[s] for s in size])
- # import IPython; IPython.embed()
+ size = relax.op.tensor_to_shape(relax.op.concat([self.env[s] for s in size]))
+ print(type(size))
if method.startswith("nearest"):
method = "nearest_neighbor"
@@ -1426,6 +1438,9 @@ class TorchFXImporter:
"sqrt": self._sqrt,
"round": self._round,
"lt": self._lt,
+ "le": self._le,
+ "gt": self._gt,
+ "ge": self._ge,
"eq": self._eq,
"ne": self._ne,
"truediv": self._truediv,
diff --git a/python/tvm/relax/utils.py b/python/tvm/relax/utils.py
index 02dd941080..8593af7b16 100644
--- a/python/tvm/relax/utils.py
+++ b/python/tvm/relax/utils.py
@@ -368,7 +368,7 @@ def gen_call_tir_inputs(
if isinstance(arg.struct_info, ShapeStructInfo):
assert isinstance(
arg, ShapeExpr
- ), "For Expr having ShapeStructInfo, emit_te now only supports ShapeExpr"
+ ), f"For Expr having ShapeStructInfo, emit_te now only supports ShapeExpr (got {type(arg)}). You may need to run tvm.relax.transform.DecomposeOpsForInference first."
return [_convert_te_arg_helper(val) for val in arg.values]
if isinstance(arg.struct_info, PrimStructInfo):
return arg.value
diff --git a/src/relax/ir/emit_te.cc b/src/relax/ir/emit_te.cc
index bfb5896c99..23b840b8ca 100644
--- a/src/relax/ir/emit_te.cc
+++ b/src/relax/ir/emit_te.cc
@@ -60,7 +60,7 @@ te::Tensor TETensor(Expr value, Map<tir::Var, PrimExpr> tir_var_map, std::string
}
ICHECK(value->struct_info_.defined()) << "value must be normalized and contain StructInfo";
auto* tensor_sinfo = GetStructInfoAs<TensorStructInfoNode>(value);
- ICHECK(tensor_sinfo) << "Value must be a tensor";
+ ICHECK(tensor_sinfo) << "Value must be a tensor but it is a " << value->GetTypeKey();
auto* shape_expr = tensor_sinfo->shape.as<ShapeExprNode>();
CHECK(shape_expr)
<< "ValueError: Expression does not have an known symbolic shape, please consider use "
diff --git a/src/tir/analysis/estimate_flops.cc b/src/tir/analysis/estimate_flops.cc
index 1c427e5fd9..59274f9c45 100644
--- a/src/tir/analysis/estimate_flops.cc
+++ b/src/tir/analysis/estimate_flops.cc
@@ -69,6 +69,13 @@ struct TResult {
return *this;
}
+ TResult SetInvalid() {
+ for (auto& kv : data_) {
+ kv.second = -1;
+ }
+ return *this;
+ }
+
TResult MaxWith(const TResult& rhs) {
for (const auto& kv : rhs.data_) {
double& v = data_[kv.first];
@@ -140,9 +147,11 @@ class FlopEstimator : private ExprFunctor<TResult(const PrimExpr& n)>,
TResult VisitStmt_(const ForNode* loop) override {
TResult result = VisitStmt(loop->body);
const auto* int_imm = loop->extent.as<IntImmNode>();
- ICHECK(int_imm) << "TypeError: Expect the extent of a loop to be IntImm, but gets: "
- << loop->extent->GetTypeKey();
- result *= int_imm->value;
+ if(int_imm == nullptr) {
+ result.SetInvalid();
+ } else {
+ result *= int_imm->value;
+ }
return result;
}