update to cogent lab from leabra compatible branch

README.md

@@ -111,7 +111,7 @@ Here are the other packages from [Cogent Core](https://github.com/cogentcore/cor
 
 * [envs](https://github.com/emer/envs) has misc standalone environments that can be good starting points, including managing files, visual images, etc.
 
-* [ttail](https://github.com/cogentcore/core/tree/main/tensor/cmd/ttail) is a `tail` program for interactively viewing tabular (csv, tsv, etc) log files in a terminal CLI environment!  `go install cogentcore.org/core/tensor/cmd/ttail@latest` from anywhere to install.
+* [ttail](https://github.com/cogentcore/core/tree/main/tensor/cmd/ttail) is a `tail` program for interactively viewing tabular (csv, tsv, etc) log files in a terminal CLI environment!  `go install cogentcore.org/lab/tensor/cmd/ttail@latest` from anywhere to install.
 
 * [eTorch](https://github.com/emer/etorch) is the emergent interface to PyTorch models, providing emergent GUI NetView etc for these models.
 

actrf/actrf.go

@@ -7,8 +7,10 @@ package actrf
 //go:generate core generate -add-types
 
 import (
-	"cogentcore.org/core/tensor"
-	"cogentcore.org/core/tensor/stats/norm"
+	"slices"
+
+	"cogentcore.org/lab/stats/stats"
+	"cogentcore.org/lab/tensor"
 )
 
 // RF is used for computing an activation-based receptive field.
@@ -59,17 +61,15 @@ func (af *RF) InitShape(act, src tensor.Tensor) []int {
 	aNy, aNx, _, _ := tensor.Projection2DShape(act.Shape(), false)
 	sNy, sNx, _, _ := tensor.Projection2DShape(src.Shape(), false)
 	oshp := []int{aNy, aNx, sNy, sNx}
-	if tensor.EqualInts(af.RF.Shp.Sizes, oshp) {
+	if slices.Equal(af.RF.Shape().Sizes, oshp) {
 		return oshp
 	}
-	snm := []string{"ActY", "ActX", "SrcY", "SrcX"}
 	sshp := []int{sNy, sNx}
-	ssnm := []string{"SrcY", "SrcX"}
-	af.RF.SetShape(oshp, snm...)
-	af.NormRF.SetShape(oshp, snm...)
-	af.SumProd.SetShape(oshp, snm...)
-	af.NormSrc.SetShape(sshp, ssnm...)
-	af.SumSrc.SetShape(sshp, ssnm...)
+	af.RF.SetShapeSizes(oshp...)
+	af.NormRF.SetShapeSizes(oshp...)
+	af.SumProd.SetShapeSizes(oshp...)
+	af.NormSrc.SetShapeSizes(sshp...)
+	af.SumSrc.SetShapeSizes(sshp...)
 
 	af.ConfigView(&af.RF)
 	af.ConfigView(&af.NormRF)
@@ -81,10 +81,11 @@ func (af *RF) InitShape(act, src tensor.Tensor) []int {
 
 // ConfigView configures the view params on the tensor
 func (af *RF) ConfigView(tsr *tensor.Float32) {
-	tsr.SetMetaData("colormap", "Viridis")
-	tsr.SetMetaData("grid-fill", "1") // remove extra lines
-	tsr.SetMetaData("fix-min", "true")
-	tsr.SetMetaData("min", "0")
+	// todo:meta
+	// tsr.SetMetaData("colormap", "Viridis")
+	// tsr.SetMetaData("grid-fill", "1") // remove extra lines
+	// tsr.SetMetaData("fix-min", "true")
+	// tsr.SetMetaData("min", "0")
 }
 
 // Reset reinitializes the Sum accumulators -- must have called Init first
@@ -106,11 +107,11 @@ func (af *RF) Add(act, src tensor.Tensor, thr float32) {
 			if tv < thr {
 				continue
 			}
-			af.SumSrc.AddScalar([]int{sy, sx}, float64(tv))
+			af.SumSrc.SetAdd(tv, sy, sx)
 			for ay := 0; ay < aNy; ay++ {
 				for ax := 0; ax < aNx; ax++ {
 					av := float32(tensor.Projection2DValue(act, false, ay, ax))
-					af.SumProd.AddScalar([]int{ay, ax, sy, sx}, float64(av*tv))
+					af.SumProd.SetAdd(av*tv, ay, ax, sy, sx)
 				}
 			}
 		}
@@ -126,7 +127,7 @@ func (af *RF) Avg() {
 	var maxSrc float32
 	for sy := 0; sy < sNy; sy++ {
 		for sx := 0; sx < sNx; sx++ {
-			src := af.SumSrc.Value([]int{sy, sx})
+			src := af.SumSrc.Value(sy, sx)
 			if src == 0 {
 				continue
 			}
@@ -135,7 +136,7 @@ func (af *RF) Avg() {
 			}
 			for ay := 0; ay < aNy; ay++ {
 				for ax := 0; ax < aNx; ax++ {
-					oo := af.SumProd.Shape().Offset([]int{ay, ax, sy, sx})
+					oo := af.SumProd.Shape().IndexTo1D(ay, ax, sy, sx)
 					af.RF.Values[oo] = af.SumProd.Values[oo] / src
 				}
 			}
@@ -151,8 +152,7 @@ func (af *RF) Avg() {
 
 // Norm computes unit norm of RF values -- must be called after Avg
 func (af *RF) Norm() {
-	af.NormRF.CopyFrom(&af.RF)
-	norm.TensorUnit(&af.NormRF, 2) // 2 = norm within outer 2 dims = norm each src within
+	stats.UnitNormOut(&af.RF, &af.NormRF)
 }
 
 // AvgNorm computes RF as SumProd / SumTarg and then does Norm.

actrf/actrfs.go

@@ -8,7 +8,7 @@ import (
 	"fmt"
 
 	"cogentcore.org/core/base/errors"
-	"cogentcore.org/core/tensor"
+	"cogentcore.org/lab/tensor"
 )
 
 // RFs manages multiple named RF's -- each one must be initialized first

actrf/mpi.go

@@ -5,8 +5,8 @@
 package actrf
 
 import (
-	"cogentcore.org/core/base/mpi"
-	"cogentcore.org/core/tensor/tensormpi"
+	"cogentcore.org/lab/base/mpi"
+	"cogentcore.org/lab/tensor/tensormpi"
 )
 
 // MPISum aggregates RF Sum data across all processors in given mpi communicator.

actrf/running.go

@@ -4,7 +4,7 @@
 
 package actrf
 
-import "cogentcore.org/core/tensor"
+import "cogentcore.org/lab/tensor"
 
 // RunningAvg computes a running-average activation-based receptive field
 // for activities act relative to source activations src (the thing we're projecting rf onto)
@@ -17,15 +17,14 @@ func RunningAvg(out *tensor.Float32, act, src tensor.Tensor, tau float32) {
 	aNy, aNx, _, _ := tensor.Projection2DShape(act.Shape(), false)
 	tNy, tNx, _, _ := tensor.Projection2DShape(src.Shape(), false)
 	oshp := []int{aNy, aNx, tNy, tNx}
-	out.SetShape(oshp, "ActY", "ActX", "SrcY", "SrcX")
+	out.SetShapeSizes(oshp...)
 	for ay := 0; ay < aNy; ay++ {
 		for ax := 0; ax < aNx; ax++ {
 			av := float32(tensor.Projection2DValue(act, false, ay, ax))
 			for ty := 0; ty < tNy; ty++ {
 				for tx := 0; tx < tNx; tx++ {
 					tv := float32(tensor.Projection2DValue(src, false, ty, tx))
-					oi := []int{ay, ax, ty, tx}
-					oo := out.Shape().Offset(oi)
+					oo := out.Shape().IndexTo1D(ay, ax, ty, tx)
 					ov := out.Values[oo]
 					nv := cdt*ov + dt*tv*av
 					out.Values[oo] = nv

chem/stater.go

@@ -4,7 +4,7 @@
 
 package chem
 
-import "cogentcore.org/core/tensor/table"
+import "cogentcore.org/lab/table"
 
 // The Stater interface defines the functions implemented for State
 // structures containing chem state variables.

confusion/confusion.go

@@ -6,14 +6,7 @@ package confusion
 
 //go:generate core generate -add-types
 
-import (
-	"fmt"
-	"math"
-
-	"cogentcore.org/core/core"
-	"cogentcore.org/core/tensor"
-	"cogentcore.org/core/tensor/stats/simat"
-)
+/*
 
 // Matrix computes the confusion matrix, with rows representing
 // the ground truth correct class, and columns representing the
@@ -208,30 +201,4 @@ func (cm *Matrix) SaveCSV(fname core.Filename) {
 func (cm *Matrix) OpenCSV(fname core.Filename) {
 	tensor.OpenCSV(&cm.Prob, fname, ',')
 }
-
-/*
-var MatrixProps = tree.Props{
-	"ToolBar": tree.PropSlice{
-		{"SaveCSV", tree.Props{
-			"label": "Save CSV...",
-			"icon":  "file-save",
-			"desc":  "Save CSV-formatted confusion probabilities (Probs)",
-			"Args": tree.PropSlice{
-				{"CSV File Name", tree.Props{
-					"ext": ".csv",
-				}},
-			},
-		}},
-		{"OpenCSV", tree.Props{
-			"label": "Open CSV...",
-			"icon":  "file-open",
-			"desc":  "Open CSV-formatted confusion probabilities (Probs)",
-			"Args": tree.PropSlice{
-				{"Weights File Name", tree.Props{
-					"ext": ".csv",
-				}},
-			},
-		}},
-	},
-}
 */

decoder/linear.go

@@ -9,9 +9,9 @@ package decoder
 import (
 	"fmt"
 
-	"cogentcore.org/core/base/mpi"
 	"cogentcore.org/core/math32"
-	"cogentcore.org/core/tensor"
+	"cogentcore.org/lab/base/mpi"
+	"cogentcore.org/lab/tensor"
 )
 
 type ActivationFunc func(float32) float32
@@ -61,7 +61,7 @@ type Linear struct {
 // Layer is the subset of emer.Layer that is used by this code
 type Layer interface {
 	Name() string
-	UnitValuesTensor(tsr tensor.Tensor, varNm string, di int) error
+	UnitValuesTensor(tsr tensor.Values, varNm string, di int) error
 	Shape() *tensor.Shape
 }
 
@@ -111,7 +111,7 @@ func (dec *Linear) Init(nOutputs, nInputs int, poolIndex int, activationFn Activ
 	dec.NOutputs = nOutputs
 	dec.Units = make([]LinearUnit, dec.NOutputs)
 	dec.Inputs = make([]float32, dec.NInputs)
-	dec.Weights.SetShape([]int{dec.NOutputs, dec.NInputs}, "Outputs", "Inputs")
+	dec.Weights.SetShapeSizes(dec.NOutputs, dec.NInputs)
 	for i := range dec.Weights.Values {
 		dec.Weights.Values[i] = 0.1
 	}
@@ -207,7 +207,7 @@ func (dec *Linear) Input(varNm string, di int) {
 			shape := ly.Shape()
 			y := dec.PoolIndex / shape.DimSize(1)
 			x := dec.PoolIndex % shape.DimSize(1)
-			tsr = tsr.SubSpace([]int{y, x}).(*tensor.Float32)
+			tsr = tsr.SubSpace(y, x).(*tensor.Float32)
 		}
 		for j, v := range tsr.Values {
 			dec.Inputs[off+j] = v
@@ -259,7 +259,7 @@ func (dec *Linear) Back() float32 {
 // Returns SSE (sum squared error) of difference between targets and outputs.
 func (dec *Linear) BackMPI() float32 {
 	if dec.MPIDWts.Len() != dec.Weights.Len() {
-		dec.MPIDWts.CopyShapeFrom(&dec.Weights)
+		tensor.SetShapeFrom(&dec.MPIDWts, &dec.Weights)
 	}
 	var sse float32
 	for ui := range dec.Units {

decoder/linear_test.go

@@ -8,25 +8,25 @@ import (
 	"fmt"
 	"testing"
 
-	"cogentcore.org/core/tensor"
+	"cogentcore.org/lab/tensor"
 	"github.com/stretchr/testify/assert"
 )
 
 // TestLayer implements a Layer
 type TestLayer struct {
-	tensors map[string]tensor.Tensor
+	tensors map[string]tensor.Values
 }
 
 func (tl *TestLayer) Name() string {
 	return "TestLayer"
 }
 
-func (tl *TestLayer) UnitValuesTensor(tsr tensor.Tensor, varNm string, di int) error {
+func (tl *TestLayer) UnitValuesTensor(tsr tensor.Values, varNm string, di int) error {
 	src, ok := tl.tensors[varNm]
 	if !ok {
 		return fmt.Errorf("bad key: %s", varNm)
 	}
-	tsr.CopyShapeFrom(src)
+	tensor.SetShapeFrom(tsr, src)
 	tsr.CopyFrom(src)
 	return nil
 }
@@ -94,58 +94,58 @@ func TestLinearLogistic(t *testing.T) {
 
 func TestInputPool1D(t *testing.T) {
 	dec := Linear{}
-	shape := tensor.NewShape([]int{1, 5, 6, 6})
+	shape := tensor.NewShape(1, 5, 6, 6)
 	vals := make([]float32, shape.Len())
 	for i := range vals {
 		vals[i] = float32(i)
 	}
-	tsr := tensor.NewFloat32(shape.Sizes)
+	tsr := tensor.NewFloat32(shape.Sizes...)
 	tsr.SetNumRows(1)
 	for i := range tsr.Values {
 		tsr.Values[i] = vals[i]
 	}
-	layer := TestLayer{tensors: map[string]tensor.Tensor{"var0": tsr}}
+	layer := TestLayer{tensors: map[string]tensor.Values{"var0": tsr}}
 	dec.InitPool(2, &layer, 0, IdentityFunc)
 	dec.Input("var0", 0)
-	expected := tsr.SubSpace([]int{0, 0}).(*tensor.Float32).Values
+	expected := tsr.SubSpace(0, 0).(*tensor.Float32).Values
 	assert.Equal(t, expected, dec.Inputs)
 
 	dec.InitPool(2, &layer, 1, IdentityFunc)
 	dec.Input("var0", 0)
-	expected = tsr.SubSpace([]int{0, 1}).(*tensor.Float32).Values
+	expected = tsr.SubSpace(0, 1).(*tensor.Float32).Values
 	assert.Equal(t, expected, dec.Inputs)
 }
 
 func TestInputPool2D(t *testing.T) {
 	dec := Linear{}
-	shape := tensor.NewShape([]int{2, 5, 6, 6})
+	shape := tensor.NewShape(2, 5, 6, 6)
 	vals := make([]float32, shape.Len())
 	for i := range vals {
 		vals[i] = float32(i)
 	}
-	tsr := tensor.NewFloat32(shape.Sizes)
+	tsr := tensor.NewFloat32(shape.Sizes...)
 	for i := range tsr.Values {
 		tsr.Values[i] = vals[i]
 	}
 
-	layer := TestLayer{tensors: map[string]tensor.Tensor{"var0": tsr}}
+	layer := TestLayer{tensors: map[string]tensor.Values{"var0": tsr}}
 	dec.InitPool(2, &layer, 0, IdentityFunc)
 	dec.Input("var0", 0)
-	expected := tsr.SubSpace([]int{0, 0}).(*tensor.Float32).Values
+	expected := tsr.SubSpace(0, 0).(*tensor.Float32).Values
 	assert.Equal(t, expected, dec.Inputs)
 
 	dec.InitPool(2, &layer, 1, IdentityFunc)
 	dec.Input("var0", 0)
-	expected = tsr.SubSpace([]int{0, 1}).(*tensor.Float32).Values
+	expected = tsr.SubSpace(0, 1).(*tensor.Float32).Values
 	assert.Equal(t, expected, dec.Inputs)
 
 	dec.InitPool(2, &layer, 5, IdentityFunc)
 	dec.Input("var0", 0)
-	expected = tsr.SubSpace([]int{1, 0}).(*tensor.Float32).Values
+	expected = tsr.SubSpace(1, 0).(*tensor.Float32).Values
 	assert.Equal(t, expected, dec.Inputs)
 
 	dec.InitPool(2, &layer, 9, IdentityFunc)
 	dec.Input("var0", 0)
-	expected = tsr.SubSpace([]int{1, 4}).(*tensor.Float32).Values
+	expected = tsr.SubSpace(1, 4).(*tensor.Float32).Values
 	assert.Equal(t, expected, dec.Inputs)
 }

decoder/softmax.go

@@ -15,9 +15,9 @@ import (
 	"path/filepath"
 	"sort"
 
-	"cogentcore.org/core/base/mpi"
 	"cogentcore.org/core/math32"
-	"cogentcore.org/core/tensor"
+	"cogentcore.org/lab/base/mpi"
+	"cogentcore.org/lab/tensor"
 	"github.com/emer/emergent/v2/emer"
 )
 
@@ -93,7 +93,7 @@ func (sm *SoftMax) Init(ncats, ninputs int) {
 	sm.Units = make([]SoftMaxUnit, ncats)
 	sm.Sorted = make([]int, ncats)
 	sm.Inputs = make([]float32, sm.NInputs)
-	sm.Weights.SetShape([]int{sm.NCats, sm.NInputs}, "Cats", "Inputs")
+	sm.Weights.SetShapeSizes(sm.NCats, sm.NInputs)
 	for i := range sm.Weights.Values {
 		sm.Weights.Values[i] = .1
 	}
@@ -215,7 +215,7 @@ func (sm *SoftMax) Back() {
 // MPI version shares weight changes across nodes
 func (sm *SoftMax) BackMPI() {
 	if sm.MPIDWts.Len() != sm.Weights.Len() {
-		sm.MPIDWts.CopyShapeFrom(&sm.Weights)
+		tensor.SetShapeFrom(&sm.MPIDWts, &sm.Weights)
 	}
 	lr := sm.Lrate
 	for ui := range sm.Units {