LLM/Ryzen AI 7 350/2026-07-09 - Revision history

Tacoball: /* Let iGPU can use large memory size */

2026-07-10T10:18:44Z

Let iGPU can use large memory size

← Older revision		Revision as of 10:18, 10 July 2026
Line 144:		Line 144:
	</pre>		</pre>

	sudo apt install vulkan-tools		sudo apt install vulkan-tools
			sudo vim /var/lib/lemonade/.cache/lemonade/config.json

	== Optimize Lemonade config ==		== Optimize Lemonade config ==

2026-07-10T09:24:31Z

Let iGPU can use large memory size

← Older revision		Revision as of 09:24, 10 July 2026
Line 143:		Line 143:
	[ 5.304939] amdgpu 0000:66:00.0: 46677M of GTT memory ready.		[ 5.304939] amdgpu 0000:66:00.0: 46677M of GTT memory ready.
	</pre>		</pre>

			sudo apt install vulkan-tools

	== Optimize Lemonade config ==		== Optimize Lemonade config ==

2026-07-10T09:19:12Z

Let iGPU can use large memory size

← Older revision		Revision as of 09:19, 10 July 2026
Line 135:		Line 135:
	sudo update-initramfs -u		sudo update-initramfs -u
	sudo reboot		sudo reboot
			sudo dmesg \| grep -iE "amdgpu.*memory"
	</syntaxhighlight>		</syntaxhighlight>

			<pre>
			[ 5.304706] amdgpu 0000:66:00.0: Trusted Memory Zone (TMZ) feature disabled as experimental (default)
			[ 5.304936] amdgpu 0000:66:00.0: 512M of VRAM memory ready
			[ 5.304939] amdgpu 0000:66:00.0: 46677M of GTT memory ready.
			</pre>

	== Optimize Lemonade config ==		== Optimize Lemonade config ==

2026-07-10T09:16:59Z

Let iGPU can use large memory size

← Older revision		Revision as of 09:16, 10 July 2026
Line 131:		Line 131:
	options amdgpu page_pool_size=25600000		options amdgpu page_pool_size=25600000
	</pre>		</pre>

			<syntaxhighlight lang="bash">
			sudo update-initramfs -u
			sudo reboot
			</syntaxhighlight>

	== Optimize Lemonade config ==		== Optimize Lemonade config ==

2026-07-10T09:13:12Z

Let iGPU can use large memory size

← Older revision		Revision as of 09:13, 10 July 2026
Line 120:		Line 120:

	=== Let iGPU can use large memory size ===		=== Let iGPU can use large memory size ===

			Default iGPU memory limit is 536.87MB in Ubuntu 26.04

	<syntaxhighlight lang="bash">		<syntaxhighlight lang="bash">
			sudo nano /etc/modprobe.d/amdgpu.conf
			</syntaxhighlight>

	</~~syntaxhighlight~~>		<pre>
			options amdgpu ttm_pages_limit=25600000
			options amdgpu page_pool_size=25600000
			</pre>

	== Optimize Lemonade config ==		== Optimize Lemonade config ==

2026-07-10T09:12:05Z

Let iGPU can use large memory size

← Older revision		Revision as of 09:12, 10 July 2026
Line 120:		Line 120:

	=== Let iGPU can use large memory size ===		=== Let iGPU can use large memory size ===

			<syntaxhighlight lang="bash">

			</syntaxhighlight>

	== Optimize Lemonade config ==		== Optimize Lemonade config ==

2026-07-10T09:11:38Z

Enable NPU and set default models

← Older revision		Revision as of 09:11, 10 July 2026
Line 118:		Line 118:
	sudo systemctl restart lemond		sudo systemctl restart lemond
	</syntaxhighlight>		</syntaxhighlight>

			=== Let iGPU can use large memory size ===

	== Optimize Lemonade config ==		== Optimize Lemonade config ==

2026-07-10T03:06:15Z

Enable NPU and set default models

← Older revision		Revision as of 03:06, 10 July 2026
Line 110:		Line 110:
	Environment="LEMONADE_MODEL_TTL=0"		Environment="LEMONADE_MODEL_TTL=0"
	# Default models		# Default models
	ExecStartPost=/usr/bin/lemonade load model --ctx-size 65535 Gemma-4-12B-it~~-MTP~~-GGUF --backend cuda		ExecStartPost=/usr/bin/lemonade load model --ctx-size 65535 Gemma-4-12B-it-GGUF --backend cuda
	ExecStartPost=/usr/bin/lemonade load model gpt-oss-sg-20b-FLM --backend npu		ExecStartPost=/usr/bin/lemonade load model gpt-oss-sg-20b-FLM --backend npu
	</pre>		</pre>

2026-07-09T10:26:00Z

Enable NPU and set default models

← Older revision		Revision as of 10:26, 9 July 2026
Line 110:		Line 110:
	Environment="LEMONADE_MODEL_TTL=0"		Environment="LEMONADE_MODEL_TTL=0"
	# Default models		# Default models
	ExecStartPost=/usr/bin/lemonade model ~~load~~ Gemma-4-12B-it-MTP-GGUF --backend cuda		ExecStartPost=/usr/bin/lemonade load model --ctx-size 65535 Gemma-4-12B-it-MTP-GGUF --backend cuda
	ExecStartPost=/usr/bin/lemonade model ~~load~~ gpt-oss-sg-20b-FLM --backend npu		ExecStartPost=/usr/bin/lemonade load model gpt-oss-sg-20b-FLM --backend npu
	</pre>		</pre>

2026-07-09T10:01:02Z

Fix memory limit problem

← Older revision		Revision as of 10:01, 9 July 2026
Line 95:		Line 95:
	ulimit -l		ulimit -l
	flm validate		flm validate
			</syntaxhighlight>

			=== Enable NPU and set default models ===

			<syntaxhighlight lang="bash">
	sudo systemctl edit lemond		sudo systemctl edit lemond
	</syntaxhighlight>		</syntaxhighlight>