fix some lint issues

shubhambhokare1 · Oct 29, 2024 · 0799e71 · 0799e71
1 parent 41afee6
commit 0799e71
Show file tree

Hide file tree

Showing 31 changed files with 4 additions and 1 deletion.
diff --git a/...backend/test/data/node/test_scalar_dot_product_attention_3d_attn_mask_expanded/model.onnx b/...backend/test/data/node/test_scalar_dot_product_attention_3d_attn_mask_expanded/model.onnx
diff --git a/onnx/backend/test/data/node/test_scalar_dot_product_attention_3d_causal_expanded/model.onnx b/onnx/backend/test/data/node/test_scalar_dot_product_attention_3d_causal_expanded/model.onnx
diff --git a/.../node/test_scalar_dot_product_attention_3d_diff_heads_sizes_attn_mask_expanded/model.onnx b/.../node/test_scalar_dot_product_attention_3d_diff_heads_sizes_attn_mask_expanded/model.onnx
diff --git a/...ata/node/test_scalar_dot_product_attention_3d_diff_heads_sizes_causal_expanded/model.onnx b/...ata/node/test_scalar_dot_product_attention_3d_diff_heads_sizes_causal_expanded/model.onnx
diff --git a/.../test/data/node/test_scalar_dot_product_attention_3d_diff_heads_sizes_expanded/model.onnx b/.../test/data/node/test_scalar_dot_product_attention_3d_diff_heads_sizes_expanded/model.onnx
diff --git a/...ata/node/test_scalar_dot_product_attention_3d_diff_heads_sizes_scaled_expanded/model.onnx b/...ata/node/test_scalar_dot_product_attention_3d_diff_heads_sizes_scaled_expanded/model.onnx
diff --git a/...test_scalar_dot_product_attention_3d_diff_heads_with_past_and_present_expanded/model.onnx b/...test_scalar_dot_product_attention_3d_diff_heads_with_past_and_present_expanded/model.onnx
diff --git a/onnx/backend/test/data/node/test_scalar_dot_product_attention_3d_expanded/model.onnx b/onnx/backend/test/data/node/test_scalar_dot_product_attention_3d_expanded/model.onnx
diff --git a/...end/test/data/node/test_scalar_dot_product_attention_3d_gqa_attn_mask_expanded/model.onnx b/...end/test/data/node/test_scalar_dot_product_attention_3d_gqa_attn_mask_expanded/model.onnx
diff --git a/...ackend/test/data/node/test_scalar_dot_product_attention_3d_gqa_causal_expanded/model.onnx b/...ackend/test/data/node/test_scalar_dot_product_attention_3d_gqa_causal_expanded/model.onnx
diff --git a/onnx/backend/test/data/node/test_scalar_dot_product_attention_3d_gqa_expanded/model.onnx b/onnx/backend/test/data/node/test_scalar_dot_product_attention_3d_gqa_expanded/model.onnx
diff --git a/...ackend/test/data/node/test_scalar_dot_product_attention_3d_gqa_scaled_expanded/model.onnx b/...ackend/test/data/node/test_scalar_dot_product_attention_3d_gqa_scaled_expanded/model.onnx
diff --git a/...a/node/test_scalar_dot_product_attention_3d_gqa_with_past_and_present_expanded/model.onnx b/...a/node/test_scalar_dot_product_attention_3d_gqa_with_past_and_present_expanded/model.onnx
diff --git a/onnx/backend/test/data/node/test_scalar_dot_product_attention_3d_scaled_expanded/model.onnx b/onnx/backend/test/data/node/test_scalar_dot_product_attention_3d_scaled_expanded/model.onnx
diff --git a/.../data/node/test_scalar_dot_product_attention_3d_with_past_and_present_expanded/model.onnx b/.../data/node/test_scalar_dot_product_attention_3d_with_past_and_present_expanded/model.onnx
diff --git a/...backend/test/data/node/test_scalar_dot_product_attention_4d_attn_mask_expanded/model.onnx b/...backend/test/data/node/test_scalar_dot_product_attention_4d_attn_mask_expanded/model.onnx
diff --git a/onnx/backend/test/data/node/test_scalar_dot_product_attention_4d_causal_expanded/model.onnx b/onnx/backend/test/data/node/test_scalar_dot_product_attention_4d_causal_expanded/model.onnx
diff --git a/.../node/test_scalar_dot_product_attention_4d_diff_heads_sizes_attn_mask_expanded/model.onnx b/.../node/test_scalar_dot_product_attention_4d_diff_heads_sizes_attn_mask_expanded/model.onnx
diff --git a/...ata/node/test_scalar_dot_product_attention_4d_diff_heads_sizes_causal_expanded/model.onnx b/...ata/node/test_scalar_dot_product_attention_4d_diff_heads_sizes_causal_expanded/model.onnx
diff --git a/.../test/data/node/test_scalar_dot_product_attention_4d_diff_heads_sizes_expanded/model.onnx b/.../test/data/node/test_scalar_dot_product_attention_4d_diff_heads_sizes_expanded/model.onnx
diff --git a/...ata/node/test_scalar_dot_product_attention_4d_diff_heads_sizes_scaled_expanded/model.onnx b/...ata/node/test_scalar_dot_product_attention_4d_diff_heads_sizes_scaled_expanded/model.onnx
diff --git a/...test_scalar_dot_product_attention_4d_diff_heads_with_past_and_present_expanded/model.onnx b/...test_scalar_dot_product_attention_4d_diff_heads_with_past_and_present_expanded/model.onnx
diff --git a/onnx/backend/test/data/node/test_scalar_dot_product_attention_4d_expanded/model.onnx b/onnx/backend/test/data/node/test_scalar_dot_product_attention_4d_expanded/model.onnx
diff --git a/...end/test/data/node/test_scalar_dot_product_attention_4d_gqa_attn_mask_expanded/model.onnx b/...end/test/data/node/test_scalar_dot_product_attention_4d_gqa_attn_mask_expanded/model.onnx
diff --git a/...ackend/test/data/node/test_scalar_dot_product_attention_4d_gqa_causal_expanded/model.onnx b/...ackend/test/data/node/test_scalar_dot_product_attention_4d_gqa_causal_expanded/model.onnx
diff --git a/onnx/backend/test/data/node/test_scalar_dot_product_attention_4d_gqa_expanded/model.onnx b/onnx/backend/test/data/node/test_scalar_dot_product_attention_4d_gqa_expanded/model.onnx
diff --git a/...ackend/test/data/node/test_scalar_dot_product_attention_4d_gqa_scaled_expanded/model.onnx b/...ackend/test/data/node/test_scalar_dot_product_attention_4d_gqa_scaled_expanded/model.onnx
diff --git a/...a/node/test_scalar_dot_product_attention_4d_gqa_with_past_and_present_expanded/model.onnx b/...a/node/test_scalar_dot_product_attention_4d_gqa_with_past_and_present_expanded/model.onnx
diff --git a/onnx/backend/test/data/node/test_scalar_dot_product_attention_4d_scaled_expanded/model.onnx b/onnx/backend/test/data/node/test_scalar_dot_product_attention_4d_scaled_expanded/model.onnx
diff --git a/.../data/node/test_scalar_dot_product_attention_4d_with_past_and_present_expanded/model.onnx b/.../data/node/test_scalar_dot_product_attention_4d_with_past_and_present_expanded/model.onnx
diff --git a/onnx/defs/nn/defs.cc b/onnx/defs/nn/defs.cc
@@ -4,6 +4,7 @@
 
 #include <algorithm>
 #include <cmath>
+#include <limits>
 
 #include "onnx/common/assertions.h"
 #include "onnx/defs/function.h"
@@ -3167,9 +3168,10 @@ ONNX_OPERATOR_SET_SCHEMA(
             // An error is thrown if both attn_mask and is_causal are set.
             auto* is_causal_attr = ctx.getAttribute("is_causal");
             int64_t is_causal = (is_causal_attr != nullptr) ? is_causal_attr->i() : 0;
+            float neg_inf = -std::numeric_limits<float>::infinity();
             builder.Add("TempMask = ConstantOfShape(AttnBiasShape)", "value", mktensor(1))
                 .Add("TempMaskTri = Trilu <upper = 0> (TempMask, Zero1D)")
-                .Const1D("FloatInf", static_cast<float>('-inf'))
+                .Const1D("FloatInf", neg_inf)
                 .Add("CasualAttnBias = Where(TempMaskTri, AttnBiasZeros, FloatInf)")
                 .Const("IsCasual", is_causal)
                 .Add("AttnBias = Where(IsCausal, CasualAttnBias, AttnBiasZeros)")
@@ -3184,6 +3186,7 @@ ONNX_OPERATOR_SET_SCHEMA(
               .Add("GQACond1 = Not(NGQACond1)")
               .Add("DivNumHeads = Div(QNumHeads, KVNumHeads)")
               .Add("IDivNumHeads = Cast(DivNumHeads)", "to", int_type)
+              .Add("RemainderNumHeads = Mod(QNumHeads, KVNumHeads)")
               .Add("GQACond2 = Equal(RemainderNumHeads, Zero1D)")
               .Add("GQACond = And(GQACond1, GQACond2)")
               .Add("InterleaveShape = Concat <axis = 0> (One1D, IDivNumHeads, One1D, One1D)")