Merge branch 'ollama:main' into arm64static

.github/ISSUE_TEMPLATE/10_bug_report.yml

@@ -0,0 +1,60 @@
+name: Bug report
+labels: [bug]
+description: Something isn't working right.
+body:
+  - type: textarea
+    id: description
+    attributes:
+      label: What is the issue?
+      description: What happened? What did you expect to happen?
+    validations:
+      required: true
+  - type: dropdown
+    id: os
+    attributes:
+      label: OS
+      description: Which operating system are you using?
+      multiple: true
+      options:
+        - Linux
+        - macOS
+        - Windows
+        - Docker
+        - WSL2
+    validations:
+      required: false
+  - type: dropdown
+    id: gpu
+    attributes:
+      label: GPU
+      description: Which GPU are you using?
+      multiple: true
+      options:
+        - Nvidia
+        - AMD
+        - Intel
+        - Apple
+        - Other
+    validations:
+      required: false
+  - type: dropdown
+    id: cpu
+    attributes:
+      label: CPU
+      description: Which CPU are you using?
+      multiple: true
+      options:
+        - Intel
+        - AMD
+        - Apple
+        - Other
+    validations:
+      required: false
+  - type: input
+    id: version
+    attributes:
+      label: Ollama version
+      description: What version of Ollama are you using? (`ollama --version`)
+      placeholder: e.g., 0.1.32
+    validations:
+      required: false

.github/ISSUE_TEMPLATE/10_model_request.yml

@@ -1,18 +0,0 @@
-name: Model request
-description: Request a new model for the library
-labels: [mr]
-body:
-  - type: markdown
-    attributes:
-      value: |
-        Please check if your Model request is [already available](https://ollama.com/search) or that you cannot [import it](https://github.com/ollama/ollama/blob/main/docs/import.md#import-a-model) yourself.
-        Tell us about which Model you'd like to see in the library!
-  - type: textarea
-    id: problem
-    attributes:
-      label: What model would you like?
-      description: Please provide a link to the model.
-  - type: markdown
-    attributes:
-      value: |
-        Thanks for filing a model request!

.github/ISSUE_TEMPLATE/20_feature_request.yml

@@ -1,41 +1,11 @@
 name: Feature request
-description: Propose a new feature
-labels: [needs-triage, fr]
+labels: ['feature request']
+description: Request a new feature.
 body:
-  - type: markdown
-    attributes:
-      value: |
-        Please check if your feature request is [already filed](https://github.com/ollama/ollama/issues).
-        Tell us about your idea!
   - type: textarea
     id: problem
     attributes:
-      label: What are you trying to do?
+      label: What new feature would you like to see?
       description: Tell us about the problem you're trying to solve.
     validations:
       required: false
-  - type: textarea
-    id: solution
-    attributes:
-      label: How should we solve this?
-      description: If you have an idea of how you'd like to see this feature work, let us know.
-    validations:
-      required: false
-  - type: textarea
-    id: alternative
-    attributes:
-      label: What is the impact of not solving this?
-      description: (How) Are you currently working around the issue?
-    validations:
-      required: false
-  - type: textarea
-    id: context
-    attributes:
-      label: Anything else?
-      description: Any additional context to share, e.g., links
-    validations:
-      required: false
-  - type: markdown
-    attributes:
-      value: |
-        Thanks for filing a feature request!

.github/ISSUE_TEMPLATE/30_model_request.yml

@@ -0,0 +1,9 @@
+name: Model request
+labels: ['model request']
+description: Request a new model.
+body:
+  - type: textarea
+    id: problem
+    attributes:
+      label: What model would you like?
+      description: Please provide a link to the model.

.github/ISSUE_TEMPLATE/90_bug_report.yml

@@ -1,125 +0,0 @@
-name: Bug report
-description: File a bug report. If you need help, please join our Discord server.
-labels: [needs-triage, bug]
-body:
-  - type: markdown
-    attributes:
-      value: |
-        Please check if your bug is [already filed](https://github.com/ollama/ollama/issues) before filing a new one.
-  - type: textarea
-    id: what-happened
-    attributes:
-      label: What is the issue?
-      description: What happened? What did you expect to happen?
-    validations:
-      required: true
-  - type: textarea
-    id: what-was-expected
-    attributes:
-      label: What did you expect to see?
-      description: What did you expect to see/happen instead?
-    validations:
-      required: false
-  - type: textarea
-    id: steps
-    attributes:
-      label: Steps to reproduce
-      description: What are the steps you took that hit this issue?
-    validations:
-      required: false
-  - type: textarea
-    id: changes
-    attributes:
-      label: Are there any recent changes that introduced the issue?
-      description: If so, what are those changes?
-    validations:
-      required: false
-  - type: dropdown
-    id: os
-    attributes:
-      label: OS
-      description: What OS are you using? You may select more than one.
-      multiple: true
-      options:
-        - Linux
-        - macOS
-        - Windows
-        - Other
-    validations:
-      required: false
-  - type: dropdown
-    id: architecture
-    attributes:
-      label: Architecture
-      description: What architecture are you using? You may select more than one.
-      multiple: true
-      options:
-        - arm64
-        - amd64
-        - x86
-        - Other
-  - type: dropdown
-    id: platform
-    attributes:
-      label: Platform
-      description: What platform are you using? You may select more than one.
-      multiple: true
-      options:
-        - Docker
-        - WSL
-        - WSL2
-    validations:
-      required: false
-  - type: input
-    id: ollama-version
-    attributes:
-      label: Ollama version
-      description: What Ollama version are you using? (`ollama --version`)
-      placeholder: e.g., 1.14.4
-    validations:
-      required: false
-  - type: dropdown
-    id: gpu
-    attributes:
-      label: GPU
-      description: What GPU, if any, are you using? You may select more than one.
-      multiple: true
-      options:
-        - Nvidia
-        - AMD
-        - Intel
-        - Apple
-        - Other
-    validations:
-      required: false
-  - type: textarea
-    id: gpu-info
-    attributes:
-      label: GPU info
-      description: What GPU info do you have? (`nvidia-smi`, `rocminfo`, `system_profiler SPDisplaysDataType`, etc.)
-    validations:
-      required: false
-  - type: dropdown
-    id: cpu
-    attributes:
-      label: CPU
-      description: What CPU are you using? You may select more than one.
-      multiple: true
-      options:
-        - Intel
-        - AMD
-        - Apple
-        - Other
-    validations:
-      required: false
-  - type: textarea
-    id: other-software
-    attributes:
-      label: Other software
-      description: What other software are you using that might be related to this issue?
-    validations:
-      required: false
-  - type: markdown
-    attributes:
-      value: |
-        Thanks for filing a bug report!

README.md

@@ -60,7 +60,6 @@ Here are some example models that can be downloaded:
 | Llama 2 13B        | 13B        | 7.3GB | `ollama run llama2:13b`        |
 | Llama 2 70B        | 70B        | 39GB  | `ollama run llama2:70b`        |
 | Orca Mini          | 3B         | 1.9GB | `ollama run orca-mini`         |
-| Vicuna             | 7B         | 3.8GB | `ollama run vicuna`            |
 | LLaVA              | 7B         | 4.5GB | `ollama run llava`             |
 | Gemma              | 2B         | 1.4GB | `ollama run gemma:2b`          |
 | Gemma              | 7B         | 4.8GB | `ollama run gemma:7b`          |
@@ -378,3 +377,6 @@ See the [API documentation](./docs/api.md) for all endpoints.
 - [Page Assist](https://github.com/n4ze3m/page-assist) (Chrome Extension)
 - [AI Telegram Bot](https://github.com/tusharhero/aitelegrambot) (Telegram bot using Ollama in backend)
 - [AI ST Completion](https://github.com/yaroslavyaroslav/OpenAI-sublime-text) (Sublime Text 4 AI assistant plugin with Ollama support)
+
+### Supported backends 
+- [llama.cpp](https://github.com/ggerganov/llama.cpp) project founded by Georgi Gerganov. 

llm/ggml.go

@@ -164,7 +164,8 @@ func (ts Tensors) Layers() map[string]Layer {
 	for _, t := range ts {
 		parts := strings.Split(t.Name, ".")
 		if parts[0] == "blk" {
-			parts = parts[1:]
+			// join first and second part, e.g. blk.%d
+			parts = append([]string{fmt.Sprintf("%s.%s", parts[0], parts[1])}, parts[2:]...)
 		}
 
 		if _, ok := layers[parts[0]]; !ok {

llm/server.go

@@ -97,7 +97,7 @@ func NewLlamaServer(model string, adapters, projectors []string, opts api.Option
 	var layerCount int
 	layers := ggml.Tensors().Layers()
 	for i := 0; i < int(ggml.KV().BlockCount()); i++ {
-		memoryLayer := layers[fmt.Sprintf("%d", i)].size()
+		memoryLayer := layers[fmt.Sprintf("blk.%d", i)].size()
 
 		// KV is proportional to the number of layers
 		memoryLayer += kv / ggml.KV().BlockCount()
@@ -109,7 +109,14 @@ func NewLlamaServer(model string, adapters, projectors []string, opts api.Option
 		}
 	}
 
-	memoryLayerOutput := layers["output"].size()
+	var memoryLayerOutput uint64
+	for k, v := range layers {
+		if !strings.HasPrefix(k, "blk.") {
+			slog.Info("aaa", "name", k, "size", format.HumanBytes2(v.size()))
+			memoryLayerOutput += v.size()
+		}
+	}
+
 	memoryRequiredTotal += memoryLayerOutput
 
 	if info.Library == "metal" && memoryRequiredTotal > info.TotalMemory {
@@ -124,16 +131,47 @@ func NewLlamaServer(model string, adapters, projectors []string, opts api.Option
 		opts.NumGPU = layerCount
 	}
 
+	memoryWeights := memoryRequiredTotal - memoryMinimum - graphFullOffload - kv
+
 	slog.Info(
 		"offload to gpu",
-		"reallayers", opts.NumGPU,
-		"layers", layerCount,
-		"required", format.HumanBytes2(memoryRequiredTotal),
-		"used", format.HumanBytes2(memoryRequiredPartial),
-		"available", format.HumanBytes2(memoryAvailable),
-		"kv", format.HumanBytes2(kv),
-		"fulloffload", format.HumanBytes2(graphFullOffload),
-		"partialoffload", format.HumanBytes2(graphPartialOffload),
+		slog.Group(
+			"layers",
+			// actual number of layers offloaded
+			"real", opts.NumGPU,
+			// estimated number of layers that can be offloaded
+			"estimate", layerCount,
+		),
+		slog.Group(
+			"memory",
+			// memory available for offloading
+			"available", format.HumanBytes2(memoryAvailable),
+			slog.Group(
+				"required",
+				// memory required for full offloading
+				"full", format.HumanBytes2(memoryRequiredTotal),
+				// memory required to offload layers.estimate layers
+				"partial", format.HumanBytes2(memoryRequiredPartial),
+				// memory of KV cache
+				"kv", format.HumanBytes2(kv),
+			),
+			slog.Group(
+				"weights",
+				// memory of the weights
+				"total", format.HumanBytes2(memoryWeights),
+				// memory of repeating layers
+				"repeating", format.HumanBytes2(memoryWeights-memoryLayerOutput),
+				// memory of non-repeating layers
+				"nonrepeating", format.HumanBytes2(memoryLayerOutput),
+			),
+			slog.Group(
+				"graph",
+				// memory of graph when fully offloaded
+				"full", format.HumanBytes2(graphFullOffload),
+				// memory of graph when not fully offloaded
+				"partial", format.HumanBytes2(graphPartialOffload),
+			),
+		),
 	)
 
 	if len(adapters) > 1 {