From d73cd076740db9c111d0e58ddd4486904469e75e Mon Sep 17 00:00:00 2001
From: Aarnav Pai <52203828+arnu515@users.noreply.github.com>
Date: Tue, 9 Jun 2026 23:16:27 +0530
Subject: [PATCH 1/6] graph: Fix granite speech model inference by applying
 embedding scale when deepstack is not used (#24357)

* llama-graph : apply embedding scale when deepstack is not used

* nits: remove non-existant hunyuan-vl from the tests

* apply suggestion from @gabe-l-hart

---------

Co-authored-by: Xuan Son Nguyen <son@huggingface.co>
---
 src/llama-graph.cpp | 6 +++---
 tools/mtmd/tests.sh | 1 -
 2 files changed, 3 insertions(+), 4 deletions(-)

diff --git a/src/llama-graph.cpp b/src/llama-graph.cpp
index 4cc4a4a16..3d942ba4f 100644
--- a/src/llama-graph.cpp
+++ b/src/llama-graph.cpp
@@ -1873,9 +1873,9 @@ ggml_tensor * llm_graph_context::build_inp_embd(ggml_tensor * tok_embd) const {
     res->t_inp_embd = cur;
 
     // For Granite architecture
-    // NOTE: Only apply scale to token inputs. Raw embeddings are assumed to be
-    //  multimodal inputs that should not be scaled.
-    if (ubatch.token && hparams.f_embedding_scale != 0.0f) {
+    // NOTE: For deepstack models, only apply scale to token inputs (ie text-only input).
+    //  Raw embeddings are assumed to be multimodal inputs that should not be scaled.
+    if (hparams.f_embedding_scale != 0.0f && (ubatch.token || hparams.n_deepstack_layers == 0)) {
         if (!ggml_is_contiguous(cur)) {
             cur = ggml_cont(ctx0, cur);
         }
diff --git a/tools/mtmd/tests.sh b/tools/mtmd/tests.sh
index 83416fb27..5da48d61b 100755
--- a/tools/mtmd/tests.sh
+++ b/tools/mtmd/tests.sh
@@ -91,7 +91,6 @@ add_test_vision "ggml-org/LightOnOCR-1B-1025-GGUF:Q8_0"
 add_test_vision "ggml-org/DeepSeek-OCR-GGUF:Q8_0" -p "Free OCR." --chat-template deepseek-ocr
 add_test_vision "ggml-org/dots.ocr-GGUF:Q8_0" -p "OCR"
 add_test_vision "ggml-org/HunyuanOCR-GGUF:Q8_0" -p "OCR"
-add_test_vision "ggml-org/HunyuanVL-4B-GGUF:Q8_0"
 add_test_vision "ggml-org/gemma-4-E2B-it-GGUF:Q8_0" --jinja
 
 add_test_audio  "ggml-org/ultravox-v0_5-llama-3_2-1b-GGUF:Q8_0"

From 76da2450a4f2cc9ce6c7fc8229e25dc0a4b41e5d Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?R=C3=A9my=20Mathieu?= <remeh@remeh.fr>
Date: Tue, 9 Jun 2026 21:33:22 +0200
Subject: [PATCH 2/6] webui: implement pinned conversations support (#21387)

* webui: implement pinned conversations support

* webui: linter/prettier pass

* Fix the unused handleMobileSidebarItemClick from the component.

* the search should find pinned conversations as well

Co-authored-by: Pascal <admin@serveurperso.com>

---------

Co-authored-by: Pascal <admin@serveurperso.com>
---
 .../SidebarNavigation.svelte                  | 52 +++++++++++++++++--
 .../SidebarNavigationConversationItem.svelte  | 16 +++++-
 tools/ui/src/lib/services/database.service.ts | 16 ++++++
 .../ui/src/lib/stores/conversations.svelte.ts | 39 +++++++++++++-
 tools/ui/src/lib/types/database.d.ts          |  1 +
 5 files changed, 117 insertions(+), 7 deletions(-)

diff --git a/tools/ui/src/lib/components/app/navigation/SidebarNavigation/SidebarNavigation.svelte b/tools/ui/src/lib/components/app/navigation/SidebarNavigation/SidebarNavigation.svelte
index ddaf4d5b8..1fa7722f2 100644
--- a/tools/ui/src/lib/components/app/navigation/SidebarNavigation/SidebarNavigation.svelte
+++ b/tools/ui/src/lib/components/app/navigation/SidebarNavigation/SidebarNavigation.svelte
@@ -1,7 +1,7 @@
 <script lang="ts">
 	import { goto } from '$app/navigation';
 	import { page } from '$app/state';
-	import { Trash2, Pencil, X } from '@lucide/svelte';
+	import { Trash2, Pencil, Pin, X } from '@lucide/svelte';
 	import { Button } from '$lib/components/ui/button';
 	import { DialogConfirmation } from '$lib/components/app';
 	import SidebarNavigationActions from './SidebarNavigationActions.svelte';
@@ -52,6 +52,14 @@
 
 	let conversationTree = $derived(buildConversationTree(filteredConversations));
 
+	let pinnedConversations = $derived.by(() => {
+		return conversationTree.filter(({ conversation }) => conversation.pinned);
+	});
+
+	let unpinnedConversations = $derived.by(() => {
+		return conversationTree.filter(({ conversation }) => !conversation.pinned);
+	});
+
 	let selectedConversationHasDescendants = $derived.by(() => {
 		if (!selectedConversation) return false;
 
@@ -199,6 +207,41 @@
 			/>
 		</Sidebar.Header>
 
+		{#if !isSearchModeActive && pinnedConversations.length > 0}
+			<Sidebar.Group class="p-0 px-4">
+				<Sidebar.GroupLabel>
+					<div class="flex items-center gap-1">
+						<Pin class="h-3.5 w-3.5" />
+						<span>Pinned</span>
+					</div>
+				</Sidebar.GroupLabel>
+				<Sidebar.GroupContent>
+					<Sidebar.Menu>
+						{#each pinnedConversations as { conversation, depth } (conversation.id)}
+							<Sidebar.MenuItem class="mb-1 p-0">
+								<SidebarNavigationConversationItem
+									conversation={{
+										id: conversation.id,
+										name: conversation.name,
+										lastModified: conversation.lastModified,
+										currNode: conversation.currNode,
+										forkedFromConversationId: conversation.forkedFromConversationId,
+										pinned: conversation.pinned
+									}}
+									{depth}
+									isActive={currentChatId === conversation.id}
+									onSelect={selectConversation}
+									onEdit={handleEditConversation}
+									onDelete={handleDeleteConversation}
+									onStop={handleStopGeneration}
+								/>
+							</Sidebar.MenuItem>
+						{/each}
+					</Sidebar.Menu>
+				</Sidebar.GroupContent>
+			</Sidebar.Group>
+		{/if}
+
 		<Sidebar.Group class="mt-2 h-[calc(100vh-21rem)] space-y-2 p-0 px-3">
 			{#if (filteredConversations.length > 0 && isSearchModeActive) || !isSearchModeActive}
 				<Sidebar.GroupLabel>
@@ -208,7 +251,7 @@
 
 			<Sidebar.GroupContent>
 				<Sidebar.Menu>
-					{#each conversationTree as { conversation, depth } (conversation.id)}
+					{#each isSearchModeActive ? conversationTree : unpinnedConversations as { conversation, depth } (conversation.id)}
 						<Sidebar.MenuItem class="mb-1 p-0">
 							<SidebarNavigationConversationItem
 								conversation={{
@@ -216,7 +259,8 @@
 									name: conversation.name,
 									lastModified: conversation.lastModified,
 									currNode: conversation.currNode,
-									forkedFromConversationId: conversation.forkedFromConversationId
+									forkedFromConversationId: conversation.forkedFromConversationId,
+									pinned: conversation.pinned
 								}}
 								{depth}
 								isActive={currentChatId === conversation.id}
@@ -228,7 +272,7 @@
 						</Sidebar.MenuItem>
 					{/each}
 
-					{#if conversationTree.length === 0}
+					{#if (isSearchModeActive ? conversationTree : unpinnedConversations).length === 0}
 						<div class="px-2 py-4 text-center">
 							<p class="mb-4 p-4 text-sm text-muted-foreground">
 								{searchQuery.length > 0
diff --git a/tools/ui/src/lib/components/app/navigation/SidebarNavigation/SidebarNavigationConversationItem.svelte b/tools/ui/src/lib/components/app/navigation/SidebarNavigation/SidebarNavigationConversationItem.svelte
index e38a93738..b1c2b78f6 100644
--- a/tools/ui/src/lib/components/app/navigation/SidebarNavigation/SidebarNavigationConversationItem.svelte
+++ b/tools/ui/src/lib/components/app/navigation/SidebarNavigation/SidebarNavigationConversationItem.svelte
@@ -6,7 +6,9 @@
 		Download,
 		Loader2,
 		Square,
-		GitBranch
+		GitBranch,
+		Pin,
+		PinOff
 	} from '@lucide/svelte';
 	import { DropdownMenuActions } from '$lib/components/app';
 	import * as Tooltip from '$lib/components/ui/tooltip';
@@ -57,6 +59,10 @@
 		onStop?.(conversation.id);
 	}
 
+	function handleTogglePin() {
+		conversationsStore.toggleConversationPin(conversation.id);
+	}
+
 	function handleGlobalEditEvent(event: Event) {
 		const customEvent = event as CustomEvent<{ conversationId: string }>;
 
@@ -170,6 +176,14 @@
 				triggerTooltip="More actions"
 				bind:open={dropdownOpen}
 				actions={[
+					{
+						icon: conversation.pinned ? PinOff : Pin,
+						label: conversation.pinned ? 'Unpin' : 'Pin',
+						onclick: (e: Event) => {
+							e.stopPropagation();
+							handleTogglePin();
+						}
+					},
 					{
 						icon: Pencil,
 						label: 'Edit',
diff --git a/tools/ui/src/lib/services/database.service.ts b/tools/ui/src/lib/services/database.service.ts
index 457867d98..3338439bb 100644
--- a/tools/ui/src/lib/services/database.service.ts
+++ b/tools/ui/src/lib/services/database.service.ts
@@ -344,6 +344,22 @@ export class DatabaseService {
 	 *
 	 */
 
+	/**
+	 * Toggles the pinned status of a conversation.
+	 *
+	 * @param id - Conversation ID
+	 * @returns The new pinned status
+	 */
+	static async toggleConversationPin(id: string): Promise<boolean> {
+		const conversation = await db.conversations.get(id);
+		if (!conversation) {
+			throw new Error(`Conversation ${id} not found`);
+		}
+		const newPinnedState = !conversation.pinned;
+		await this.updateConversation(id, { pinned: newPinnedState });
+		return newPinnedState;
+	}
+
 	/**
 	 * Updates the conversation's current node (active branch).
 	 * This determines which conversation path is currently being viewed.
diff --git a/tools/ui/src/lib/stores/conversations.svelte.ts b/tools/ui/src/lib/stores/conversations.svelte.ts
index cab6f59fa..bba5191ed 100644
--- a/tools/ui/src/lib/stores/conversations.svelte.ts
+++ b/tools/ui/src/lib/stores/conversations.svelte.ts
@@ -506,6 +506,33 @@ class ConversationsStore {
 		}
 	}
 
+	/**
+	 * Toggles the pinned status of a conversation.
+	 * @param convId - The conversation ID to toggle
+	 * @returns The new pinned status
+	 */
+	async toggleConversationPin(convId: string): Promise<boolean> {
+		try {
+			const newPinnedState = await DatabaseService.toggleConversationPin(convId);
+
+			const convIndex = this.conversations.findIndex((c) => c.id === convId);
+
+			if (convIndex !== -1) {
+				this.conversations[convIndex].pinned = newPinnedState;
+				this.conversations = [...this.conversations];
+			}
+
+			if (this.activeConversation?.id === convId) {
+				this.activeConversation = { ...this.activeConversation, pinned: newPinnedState };
+			}
+
+			return newPinnedState;
+		} catch (error) {
+			console.error('Failed to toggle conversation pin:', error);
+			return false;
+		}
+	}
+
 	/**
 	 * Updates conversation title with optional confirmation dialog based on settings
 	 * @param convId - The conversation ID to update
@@ -1057,6 +1084,14 @@ export const isConversationsInitialized = () => conversationsStore.isInitialized
  * Builds a flat tree of conversations with depth levels for nested forks.
  * Accepts a pre-filtered list so search filtering stays in the component.
  */
+
+// Pinned conversations first, then by lastModified descending
+const comparePinnedThenRecent = (a: DatabaseConversation, b: DatabaseConversation) => {
+	if (a.pinned && !b.pinned) return -1;
+	if (!a.pinned && b.pinned) return 1;
+	return b.lastModified - a.lastModified;
+};
+
 export function buildConversationTree(convs: DatabaseConversation[]): ConversationTreeItem[] {
 	const childrenByParent = new SvelteMap<string, DatabaseConversation[]>();
 	const forkIds = new SvelteSet<string>();
@@ -1081,7 +1116,7 @@ export function buildConversationTree(convs: DatabaseConversation[]): Conversati
 
 		const children = childrenByParent.get(conv.id);
 		if (children) {
-			children.sort((a, b) => b.lastModified - a.lastModified);
+			children.sort(comparePinnedThenRecent);
 
 			for (const child of children) {
 				walk(child, depth + 1);
@@ -1089,7 +1124,7 @@ export function buildConversationTree(convs: DatabaseConversation[]): Conversati
 		}
 	}
 
-	const roots = convs.filter((c) => !forkIds.has(c.id));
+	const roots = convs.filter((c) => !forkIds.has(c.id)).sort(comparePinnedThenRecent);
 	for (const root of roots) {
 		walk(root, 0);
 	}
diff --git a/tools/ui/src/lib/types/database.d.ts b/tools/ui/src/lib/types/database.d.ts
index add628476..b01a7b964 100644
--- a/tools/ui/src/lib/types/database.d.ts
+++ b/tools/ui/src/lib/types/database.d.ts
@@ -15,6 +15,7 @@ export interface DatabaseConversation {
 	thinkingEnabled?: boolean;
 	reasoningEffort?: ReasoningEffort;
 	forkedFromConversationId?: string;
+	pinned?: boolean;
 }
 
 export interface DatabaseMessageExtraAudioFile {

From d2e22ed975e3464ff8108542c840733b488f165f Mon Sep 17 00:00:00 2001
From: ddh0 <chemist-mulches-39@icloud.com>
Date: Wed, 10 Jun 2026 02:31:35 -0500
Subject: [PATCH 3/6] speculative : fix "ngram-map-k4v" name in logging
 (#24253)

This is a non-functional change.

When using `--spec-type ngram-map-k4v`, the log messages at startup and
runtime say `ngram-map-k`. Added logic in the in the constructor of
`common_speculative_impl_ngram_map_k` to pass the correct
`COMMON_SPECULATIVE_TYPE_NGRAM_MAP_K4V` when `config.key_only` is
`false`.

After this change, the log messages use the correct name.
---
 common/speculative.cpp | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/common/speculative.cpp b/common/speculative.cpp
index 86c1e6a42..653f932c9 100644
--- a/common/speculative.cpp
+++ b/common/speculative.cpp
@@ -843,7 +843,8 @@ struct common_speculative_impl_ngram_map_k : public common_speculative_impl {
     common_speculative_impl_ngram_map_k(
             const common_ngram_map & config,
             uint32_t n_seq)
-        : common_speculative_impl(COMMON_SPECULATIVE_TYPE_NGRAM_MAP_K, n_seq)
+        : common_speculative_impl(config.key_only ? COMMON_SPECULATIVE_TYPE_NGRAM_MAP_K
+            : COMMON_SPECULATIVE_TYPE_NGRAM_MAP_K4V, n_seq)
     {
         for (uint32_t i = 0; i < n_seq; i++) {
             this->config.push_back(config);

From 039e20a2db9e87b2477c76cc04905f3e1acad77f Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Sigbj=C3=B8rn=20Skj=C3=A6ret?= <sigbjorn.skjaeret@scala.com>
Date: Wed, 10 Jun 2026 09:45:20 +0200
Subject: [PATCH 4/6] ci : bump komac version (#24396)

---
 .github/workflows/winget.yml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/.github/workflows/winget.yml b/.github/workflows/winget.yml
index 420a98f90..69e24f940 100644
--- a/.github/workflows/winget.yml
+++ b/.github/workflows/winget.yml
@@ -17,7 +17,7 @@ jobs:
 
       - name: Install komac
         run: |
-          cargo binstall komac@2.15.0 -y
+          cargo binstall komac@2.16.0 -y
 
       - name: Find latest release
         id: find_latest_release

From fb83cc9a072b4e85b329d4f95c59478180204c87 Mon Sep 17 00:00:00 2001
From: Oliver Simons <osimons@nvidia.com>
Date: Wed, 10 Jun 2026 14:27:08 +0200
Subject: [PATCH 5/6] CUDA: Fix ssm_scan_f32 data-races (#24360)

* Add missing syncthreads before resuing cub_temp_storage

__syncthreads() is required before being allowed to resue TempStorage
smem:
https://nvidia.github.io/cccl/unstable/cub/api/classcub_1_1BlockLoad.html#_CPPv4I0EN3cub9BlockLoad4LoadEv20RandomAccessIteratorRA14ItemsPerThread_1Ti

* Add one more missing __syncthreads

Could also double-buffer, but alternative is to simply ensure all
threads have read smem* before writing to it again in the next loop
iteration

* Remove unused smem from ssm_scan_f32
---
 ggml/src/ggml-cuda/ssm-scan.cu | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/ggml/src/ggml-cuda/ssm-scan.cu b/ggml/src/ggml-cuda/ssm-scan.cu
index 2e3f97c72..3022249c7 100644
--- a/ggml/src/ggml-cuda/ssm-scan.cu
+++ b/ggml/src/ggml-cuda/ssm-scan.cu
@@ -67,6 +67,7 @@ __global__ void __launch_bounds__(splitD, 1)
     __shared__ CubTempStorage cub_temp_storage;
 
     BlockLoad(cub_temp_storage.load_temp).Load(A_block, regA);
+    __syncthreads();
     BlockLoad(cub_temp_storage.load_temp).Load(s0_block, regs0);
 #else
     const int stride_s0 = src0_nb2 / sizeof(float);
@@ -105,6 +106,7 @@ __global__ void __launch_bounds__(splitD, 1)
             regs0[n] = state;
         }
         y_block[i * stride_y + threadIdx.x] = sumf;
+        __syncthreads();
     }
 
 #ifdef USE_CUB
@@ -249,9 +251,8 @@ static void ssm_scan_f32_cuda(const float * src0, const float * src1, const floa
         GGML_ASSERT(head_dim == 1);
         GGML_ASSERT(n_group == 1);
         const dim3 blocks(n_seq, (n_head + threads - 1) / threads, 1);
-        const int  smem_size = (threads * (d_state + 1) * 2) * sizeof(float);
         if (d_state == 16) {
-            const ggml_cuda_kernel_launch_params launch_params = ggml_cuda_kernel_launch_params(blocks, threads, smem_size, stream);
+            const ggml_cuda_kernel_launch_params launch_params = ggml_cuda_kernel_launch_params(blocks, threads, 0, stream);
             switch (n_tok)
             {
             case 1:

From d2462f8f7ac6d80070a587ffebf6cd73730f4280 Mon Sep 17 00:00:00 2001
From: Tarek Dakhran <tarek@liquid.ai>
Date: Wed, 10 Jun 2026 14:41:41 +0200
Subject: [PATCH 6/6] chat: fix LFM2/LFM2.5 ignoring json_schema (#24377)

The LFM2 specialized template handler only built a grammar for tool-calling,
silently ignoring json_schema from response_format.
---
 common/chat.cpp | 15 ++++++++++++---
 1 file changed, 12 insertions(+), 3 deletions(-)

diff --git a/common/chat.cpp b/common/chat.cpp
index 24e58ab06..9639af9ca 100644
--- a/common/chat.cpp
+++ b/common/chat.cpp
@@ -1647,11 +1647,12 @@ static common_chat_params common_chat_params_init_lfm2(const common_chat_templat
     data.thinking_start_tag = THINK_START;
     data.thinking_end_tag   = THINK_END;
 
-    auto has_tools         = inputs.tools.is_array() && !inputs.tools.empty();
+    auto has_tools           = inputs.tools.is_array() && !inputs.tools.empty();
+    auto has_response_format = !inputs.json_schema.is_null() && inputs.json_schema.is_object();
     // Gate by reasoning format and whether the template supports <think>
     auto extract_reasoning = inputs.reasoning_format != COMMON_REASONING_FORMAT_NONE &&
                              tmpl.source().find(THINK_START) != std::string::npos;
-    auto include_grammar   = has_tools && inputs.tool_choice != COMMON_CHAT_TOOL_CHOICE_NONE;
+    auto include_grammar   = has_response_format || (has_tools && inputs.tool_choice != COMMON_CHAT_TOOL_CHOICE_NONE);
 
     if (inputs.has_continuation()) {
         const auto & msg = inputs.continue_msg;
@@ -1674,6 +1675,10 @@ static common_chat_params common_chat_params_init_lfm2(const common_chat_templat
         }
 
         if (!has_tools || inputs.tool_choice == COMMON_CHAT_TOOL_CHOICE_NONE) {
+            if (has_response_format) {
+                auto response_format = p.content(p.schema(p.json(), "response-format-schema", inputs.json_schema));
+                return generation_prompt + reasoning + response_format + end;
+            }
             return generation_prompt + reasoning + p.content(p.rest()) + end;
         }
         auto tool_calls = p.rule("tool-calls",
@@ -1692,13 +1697,17 @@ static common_chat_params common_chat_params_init_lfm2(const common_chat_templat
     data.parser = parser.save();
 
     if (include_grammar) {
-        data.grammar_lazy = inputs.tool_choice == COMMON_CHAT_TOOL_CHOICE_AUTO;
+        data.grammar_lazy = !(has_response_format || (has_tools && inputs.tool_choice == COMMON_CHAT_TOOL_CHOICE_REQUIRED));
         data.grammar      = build_grammar([&](const common_grammar_builder & builder) {
             foreach_function(inputs.tools, [&](const json & tool) {
                 const auto & function = tool.at("function");
                 auto         schema   = function.at("parameters");
                 builder.resolve_refs(schema);
             });
+            if (has_response_format) {
+                auto schema = inputs.json_schema;
+                builder.resolve_refs(schema);
+            }
             parser.build_grammar(builder, data.grammar_lazy);
         });