PHYou do not have permission to view the full content of this post. Log in or register now.
Salamat sa tips.
-
Welcome to PHCorner Forums! Sign up for free to unlock full access and exclusive privileges, including:
- Asking questions about computer and mobile issues
- Discovering internet tips for your mobile network and computer
LM Studio users, what are your current local LLM models?
- Thread starter Draft
- Start date
Salamat sa tips.
PHC-TheGlock
Forum Master
DeepSeek v3.1
Usually ginagamit ko sya for asking about networking. Also kapag mag tratranslate EN -> CN, CN -> EN
Usually ginagamit ko sya for asking about networking. Also kapag mag tratranslate EN -> CN, CN -> EN
Usually, nandyan naman lahat. Binanggit ko lang dahil limited yung LM Studio sa LLMs. VLMs, at embeddings. It can only handle static configurations of online LLM apis or use LM Studio as an endpoint for your other apps using the local models or mixture of both. For dynamic routing of multiple apis from same or different providers, pwede yang i-partner sa litellm - from api collections. May variety na yung LLM choices for specific tasks na di kaya ng local AI (for the hardware at hand). Sa setup mo, pwede na lahat from music generation na suno grade, create your own Filipino TTS voice model, HQ video generation, atbp. sa dami ng China models na lumabas ngayon he he. Intel Core 9 na lang kulang ko with at least 64 GB at a decent RTX card or alternative na compute-only (monitor-less) GPU with 24GB VRAM kahit luma (~10K) for the AI projects...Si misis lang ang problema ko kahit akin yung pondo he he! Putris talaga itong mga "maybahay" na ito.
Dagdag mo yung api ng deepseek-3.1 dyan: You do not have permission to view the full content of this post. Log in or register now.
(Di ko na nilagyan ng invitation link for the credits.)
May free $100 to $225 free credits upon login. Good to use kung ayaw mong mag-lag pc using other tasks while online.
PHC-TheGlock
Forum Master
minsan ko lang nman sya magamit... gpu mainly for editing and rendering 3d ko 'to ginagamit
5090 user btw
Ayos naman pala yang hardware mo - nakaka-inggit he he! Marami pang spare VRAM for other tasks. Di ko area yan. Sa data and knowledge processing yung target ko using offline,online AI or mixed to cope with my hardware limitations.
Maganda ngang tanungin yang Deepseek-3.1-37B, kahit yung Qwen2.5-Coder-32B for networking analysis. Makakagawa ka ng pamalit sa E×ρréššVPN w/out spending a dime, with matching browser specifics for a specific purpose, or solve any complex networking scenarios you can think of with just a prompt - CCNA certified na sila he he. Mataas na yung proficiency nila sa networking.
Skia Monarch
Elite
Congrats din sa'yo, bossing, because of your forum, we cannot meet these Pinoy enthusiasts. Gusto ko naman maka connect ako sa mga enthusiast pagdating sa technology. Love reading this good discussion sharing their insights, knowledge and information. Mas iba talaga feeling kung human interaction. na realize ko lang ito nung napanood ko yung podcast ng Rota Wheels from Autoindustriya.com sa YouTube kung pano din sila nag meet and nag establish ng brand and daming connection through forum. Probably ka era mo sila boss draft. hehe so back to topic.
I want to hear more from you guys about taking your AI practice into using Agents and Automation. kung stabilize lang sana bansa natin and united, mas may potential pa ako makikita na maka invest tayo more sa hardware and maka on par din sa first world countries.
Skia Monarch
Elite
Nag setup ako ng gemma e2b sa low end spec ng work pc ko haha ang tagal mag results xD
Run mo dapat sa GPU, iGPU, or NPU para medyo mabilis
.PHC-TheGlock
Forum Master
Tinry ko sakin, ambilis
Ganda nya rin gamitin compare sa ibang models na nakainstall sakin.
Kung talagang low-end tulad ng akin he he, kaya mong pabilisin yan ng kaunti gamit yung ONNX model (Q4_K_M) na 3.2GB. Sa CPU mode mga 8-10 tokens/sec sa old 3rd-gen Quad ko. Pasok pa rin sa normal reading speed ng tao na 5 - 8 tokens/sec (~6 words/sec). May initial delay lang talaga bago sumagot kahit sa 4GB GPU, lalo pa sa CPU mode. May mga nabasa akong setup using gemma e2b ONNX models for ARM like Rasbberry Pi5 na 8 - 12 tokens/sec.
5090 ba ito?
Skia Monarch
Elite
yung GPU ko is GeForce 1050 Ti lang hahaRun mo dapat sa GPU, iGPU, or NPU para medyo mabilis
anong prompt at gpu boss? hahaTinry ko sakin, ambilis
Ganda nya rin gamitin compare sa ibang models na nakainstall sakin.
salamat po boss draft