minus phase fully on GPU. not getting any state back until plus phase…

… now.
emer · Nov 25, 2024 · f98952d · f98952d
1 parent 176abe6
commit f98952d
Show file tree

Hide file tree

Showing 13 changed files with 294 additions and 224 deletions.
diff --git a/axon/act-layer.go b/axon/act-layer.go
diff --git a/axon/act-layer.goal b/axon/act-layer.goal
@@ -713,6 +713,8 @@ func (ly *LayerParams) CyclePost(ctx *Context, di uint32) {
 	lpi := ly.PoolIndex(0)
 	ly.CyclePostLayer(ctx, lpi, di)
 	switch ly.Type {
+	case MatrixLayer, BGThalLayer:
+		ly.GatedFromSpkMax(ctx, di)
 	case CeMLayer:
 		ly.CyclePostCeMLayer(ctx, lpi, di)
 	case VSPatchLayer:
@@ -849,6 +851,24 @@ func (ly *LayerParams) CyclePostVSPatchLayer(ctx *Context, pi, di uint32, spi in
 
 ////////  Phase timescale
 
+// DecayStateNeuronsAll decays neural activation state by given proportion
+// (default decay values are ly.Params.Acts.Decay.Act, Glong, AHP)
+// for all data parallel indexes. Does not decay pool or layer state.
+// This is used for minus phase of Pulvinar layers to clear state in prep
+// for driver plus phase.
+func (ly *LayerParams) DecayStateNeuronsAll(ctx *Context, decay, glong, ahp float32) {
+	nn := ly.Indexes.NNeurons
+	for lni := uint32(0); lni < nn; lni++ {
+		ni := ly.Indexes.NeurSt + lni
+		if NeuronIsOff(ni) {
+			continue
+		}
+		for di := uint32(0); di < ctx.NData; di++ {
+			ly.Acts.DecayState(ctx, ni, di, decay, glong, ahp)
+		}
+	}
+}
+
 // NewStateLayer does NewState at the layer level, called 
 func (ly *LayerParams) NewStateLayer(ctx *Context) {
 	actMinusAvg := float32(0)
@@ -950,6 +970,17 @@ func (ly *LayerParams) MinusPhaseNeuron(ctx *Context, ni, di uint32) {
 	Neurons[ni, di, CaSpkPM] = Neurons[ni, di, CaSpkP]
 }
 
+// MinusPhasePost does special algorithm processing at end of minus
+func (ly *LayerParams) MinusPhasePost(ctx *Context) {
+	switch ly.Type {
+	case MatrixLayer:
+		ly.MatrixGated(ctx) // need gated state for decisions about action processing, so do in minus too
+	case PulvinarLayer:
+		ly.DecayStateNeuronsAll(ctx, 1, 1, 0)
+	default:
+	}
+}
+
 // PlusPhaseStartNeuron does neuron level plus-phase start:
 // applies Target inputs as External inputs.
 func (ly *LayerParams) PlusPhaseStartNeuron(ctx *Context, ni, di uint32) {
@@ -1195,16 +1226,6 @@ func (ly *Layer) UpdateExtFlags(ctx *Context) {
 	}
 }
 
-// MinusPhasePost does special algorithm processing at end of minus
-func (ly *Layer) MinusPhasePost(ctx *Context) {
-	switch ly.Type {
-	case MatrixLayer:
-		ly.MatrixGated(ctx) // need gated state for decisions about action processing, so do in minus too
-	case PulvinarLayer:
-		ly.DecayStateNeuronsAll(ctx, 1, 1, 0)
-	}
-}
-
 // PlusPhasePost does special algorithm processing at end of plus
 func (ly *Layer) PlusPhasePost(ctx *Context) {
 	ly.PlusPhaseActAvg(ctx)
@@ -1235,7 +1256,7 @@ func (ly *Layer) PlusPhasePost(ctx *Context) {
 	}
 	switch ly.Type {
 	case MatrixLayer:
-		ly.MatrixGated(ctx)
+		ly.Params.MatrixGated(ctx)
 	}
 }
 

diff --git a/axon/act-net.go b/axon/act-net.go
diff --git a/axon/act-net.goal b/axon/act-net.goal
@@ -33,8 +33,6 @@ func (nt *Network) Cycle(ncyc int, getNeurons bool) {
 
 	if getNeurons {
 		RunDoneLayersNeurons()
-	} else {
-		RunDoneLayers()
 	}
 
 	// todo: fix this:
@@ -99,7 +97,6 @@ func (nt *Network) ApplyExts() {
 	ctx := nt.Context()
 	nd := int(nix.NNeurons * ctx.NData)
 	RunApplyExtsNeuron(nd)
-	// note: not completed until cycle is run
 }
 
 // MinusPhase does updating after end of minus phase.
@@ -110,22 +107,7 @@ func (nt *Network) MinusPhase() {
 	pd := int(nix.NPools * ctx.NData)
 	RunMinusPhasePool(pd)
 	RunMinusPhaseNeuron(nd)
-	RunDoneLayersNeurons()
-	nt.MinusPhasePost()
-	ToGPULayersNeurons()
-	// todo:
-	// nt.GPU.SyncStateToGPU()
-}
-
-// MinusPhasePost does special CPU post processing.
-func (nt *Network) MinusPhasePost() {
-	ctx := nt.Context()
-	for _, ly := range nt.Layers {
-		if ly.Off {
-			continue
-		}
-		ly.MinusPhasePost(ctx)
-	}
+	RunMinusPhasePost(int(nix.NLayers))
 }
 
 // PlusPhaseStart does updating at the start of the plus phase:
@@ -135,7 +117,6 @@ func (nt *Network) PlusPhaseStart() {
 	ctx := nt.Context()
 	nd := int(nix.NNeurons * ctx.NData)
 	RunPlusPhaseStartNeuron(nd)
-	RunDone()
 }
 
 // PlusPhase does updating after end of plus phase
@@ -351,6 +332,12 @@ func MinusPhaseNeuron(i uint32) { //gosl:kernel
 	Layers[li].MinusPhaseNeuron(ctx, ni, di)
 }
 
+// MinusPhasePost does special algorithm post processing.
+func MinusPhasePost(li uint32) { //gosl:kernel
+	ctx := GetCtx(0)
+	Layers[li].MinusPhasePost(ctx)
+}
+
 // PlusPhaseStartNeuron is the kernel over Neurons * Data to
 // do neuron-level updating at start of plus phase.
 func PlusPhaseStartNeuron(i uint32) { //gosl:kernel

diff --git a/axon/gosl.go b/axon/gosl.go
diff --git a/axon/init-layer.go b/axon/init-layer.go
diff --git a/axon/init-layer.goal b/axon/init-layer.goal
@@ -343,22 +343,4 @@ func (ly *Layer) DecayStatePool(ctx *Context, pool int, decay, glong, ahp float3
 	}
 }
 
-// DecayStateNeuronsAll decays neural activation state by given proportion
-// (default decay values are ly.Params.Acts.Decay.Act, Glong, AHP)
-// for all data parallel indexes. Does not decay pool or layer state.
-// This is used for minus phase of Pulvinar layers to clear state in prep
-// for driver plus phase.
-func (ly *Layer) DecayStateNeuronsAll(ctx *Context, decay, glong, ahp float32) {
-	nn := ly.NNeurons
-	for lni := uint32(0); lni < nn; lni++ {
-		ni := ly.NeurStIndex + lni
-		if NeuronIsOff(ni) {
-			continue
-		}
-		for di := uint32(0); di < ctx.NData; di++ {
-			ly.Params.Acts.DecayState(ctx, ni, di, decay, glong, ahp)
-		}
-	}
-}
-