where is the llm files downloaded / stored? (https://felladrin-minisearch.hf.space)

[quantumalchemy]

Q. where is the llm files downloaded / stored? (https://felladrin-minisearch.hf.space)

[felladrin]

Hi, @quantumalchemy!

They are stored in the browser file system.

MiniSearch uses two engines for the models: Wllama and WebLLM.

Both engines use models split in small shards.

Wllama stores the files using Origin Private File System (OPFS), which you can visualize using the OPFS Explorer Chrome Extension.

WebLLM stores them using Cache Storage, which you can see via the Applications Tab in Chrome Dev Tools:

[quantumalchemy]

Thanks -- unfortunately I was using firefox and apparently doesn't show OPFS or has an exporer ext.
do you know where firefox would store these opfs files? .. cant find any ref .. guess I will ask a good ai

[felladrin]

Ah... I don't know of an extension for Firefox to visualize them, unfortunately.
But can you tell more about what is the problem you're trying to solve?

If you want to delete the files, MiniSearch has this option in the Menu to delete them all:

Or, if you'd like to download the GGUF files, they are all available on Hugging Face, where they are downloaded from. If that's the case, you can fine the list of models here:

MiniSearch/client/modules/wllama.ts

Lines 123 to 281 in 9a1b377

	export const wllamaModels: Readonly<Record<string, WllamaModel>> = {
	"smollm2-135m": {
	...createDefaultModelConfig(),
	label: "SmolLM 2 135M",
	hfRepoId: "Felladrin/gguf-Q8_0-SmolLM2-135M-Instruct",
	hfFilePath: "model.shard-00001-of-00005.gguf",
	fileSizeInMegabytes: 145,
	},
	"smollm2-360m": {
	...createDefaultModelConfig(),
	label: "SmolLM 2 360M",
	hfRepoId: "Felladrin/gguf-Q8_0-SmolLM2-360M-Instruct",
	hfFilePath: "model.shard-00001-of-00008.gguf",
	fileSizeInMegabytes: 386,
	},
	"qwen-2.5-0.5b": {
	...createDefaultModelConfig(),
	label: "Qwen 2.5 0.5B",
	hfRepoId: "Felladrin/gguf-sharded-Q4_K_S-Qwen2.5-0.5B-Instruct",
	hfFilePath: "model.shard-00001-of-00003.gguf",
	fileSizeInMegabytes: 386,
	},
	"danube-3-500m": {
	...createDefaultModelConfig(),
	label: "Danube 3 500M",
	hfRepoId: "Felladrin/gguf-sharded-Q5_K_L-h2o-danube3-500m-chat",
	hfFilePath: "model.shard-00001-of-00008.gguf",
	fileSizeInMegabytes: 547,
	},
	"amd-olmo-1b": {
	...createDefaultModelConfig(),
	label: "AMD OLMo 1B",
	hfRepoId: "Felladrin/gguf-Q5_K_L-AMD-OLMo-1B-SFT-DPO",
	hfFilePath: "model.shard-00001-of-00009.gguf",
	fileSizeInMegabytes: 872,
	},
	"granite-3.0-1b": {
	...createDefaultModelConfig(),
	label: "Granite 3.0 1B [400M]",
	hfRepoId: "Felladrin/gguf-sharded-q5_k_l-granite-3.0-1b-a400m-instruct",
	hfFilePath: "model.shard-00001-of-00019.gguf",
	fileSizeInMegabytes: 969,
	},
	"llama-3.2-1b": {
	...createDefaultModelConfig(),
	label: "Llama 3.2 1B",
	hfRepoId: "Felladrin/gguf-sharded-Q5_K_L-Llama-3.2-1B-Instruct",
	hfFilePath: "model.shard-00001-of-00005.gguf",
	fileSizeInMegabytes: 975,
	},
	"pythia-1.4b": {
	...createDefaultModelConfig(),
	label: "Pythia 1.4B",
	hfRepoId: "Felladrin/gguf-sharded-pythia-1.4b-sft-full",
	hfFilePath: "pythia-1.4b-sft-full.Q5_K_M.shard-00001-of-00011.gguf",
	fileSizeInMegabytes: 1060,
	},
	"pints-1.5b": {
	...createDefaultModelConfig(),
	label: "Pints 1.5B",
	hfRepoId: "Felladrin/gguf-sharded-Q5_K-1.5-Pints-2K-v0.1",
	hfFilePath: "model.shard-00001-of-00018.gguf",
	fileSizeInMegabytes: 1150,
	},
	"smollm2-1.7b": {
	...createDefaultModelConfig(),
	label: "SmolLM 2 1.7B",
	hfRepoId: "Felladrin/gguf-Q5_K_M-SmolLM2-1.7B-Instruct",
	hfFilePath: "model.shard-00001-of-00016.gguf",
	fileSizeInMegabytes: 1230,
	},
	"internlm-2.5-1.8b": {
	...createDefaultModelConfig(),
	label: "InternLM 2.5 1.8B",
	hfRepoId: "Felladrin/gguf-sharded-q5_k_m-internlm2_5-1_8b-chat",
	hfFilePath: "model.shard-00001-of-00009.gguf",
	fileSizeInMegabytes: 1360,
	},
	"arcee-lite": {
	...createDefaultModelConfig(),
	label: "Arcee Lite 1.5B",
	hfRepoId: "Felladrin/gguf-q5_k_l-imat-arcee-lite",
	hfFilePath: "arcee-lite-Q5_K_L.shard-00001-of-00006.gguf",
	fileSizeInMegabytes: 1430,
	},
	"granite-3.0-2b": {
	...createDefaultModelConfig(),
	label: "Granite 3.0 2B",
	hfRepoId: "Felladrin/gguf-q5_k_m-granite-3.0-2b-instruct",
	hfFilePath: "granite-3-00001-of-00023.gguf",
	fileSizeInMegabytes: 1870,
	},
	"exaone-3.5-2.4b": {
	...createDefaultModelConfig(),
	label: "EXAONE 3.5 2.4B",
	hfRepoId: "Felladrin/gguf-sharded-Q5_K_M-EXAONE-3.5-2.4B-Instruct",
	hfFilePath: "model.shard-00001-of-00009.gguf",
	fileSizeInMegabytes: 1910,
	cacheTypeV: "f16",
	},
	"gemma-2-2b": {
	...createDefaultModelConfig(),
	label: "Gemma 2 2B",
	hfRepoId: "Felladrin/gguf-sharded-gemma-2-2b-it-abliterated",
	hfFilePath: "gemma-2-2b-it-abliterated-q5_k_m-imat-00001-of-00009.gguf",
	fileSizeInMegabytes: 1920,
	},
	"llama-3.2-3b": {
	...createDefaultModelConfig(),
	label: "Llama 3.2 3B",
	hfRepoId: "Felladrin/gguf-sharded-Q5_K_L-Llama-3.2-3B-Instruct",
	hfFilePath: "model.shard-00001-of-00007.gguf",
	fileSizeInMegabytes: 2420,
	},
	"granite-3.0-3b": {
	...createDefaultModelConfig(),
	label: "Granite 3.0 3B [800M]",
	hfRepoId: "Felladrin/gguf-sharded-Q5_K_L-granite-3.0-3b-a800m-instruct",
	hfFilePath: "model.shard-00001-of-00034.gguf",
	fileSizeInMegabytes: 2450,
	},
	"minicpm3-4b": {
	...createDefaultModelConfig(),
	label: "MiniCPM 3 4B",
	hfRepoId: "Felladrin/gguf-Q4_K_M-MiniCPM3-4B",
	hfFilePath: "model.shard-00001-of-00017.gguf",
	fileSizeInMegabytes: 2470,
	contextSize: 2496,
	},
	"phi-3.5-mini-3.8b": {
	...createDefaultModelConfig(),
	label: "Phi 3.5 Mini 3.8B",
	hfRepoId: "Felladrin/gguf-q5_k_m-phi-3.5-mini-instruct",
	hfFilePath: "phi-3-00001-of-00025.gguf",
	fileSizeInMegabytes: 2820,
	contextSize: 3584,
	},
	"magpielm-4b": {
	...createDefaultModelConfig(),
	label: "MagpieLM 4B",
	hfRepoId: "Felladrin/gguf-Q5_K_M-MagpieLM-4B-Chat-v0.1",
	hfFilePath: "magpielm-4b-chat-v0-00001-of-00019.gguf",
	fileSizeInMegabytes: 3230,
	},
	"nemotron-mini-4b": {
	...createDefaultModelConfig(),
	label: "Nemotron Mini 4B",
	hfRepoId: "Felladrin/gguf-Q5_K_M-Nemotron-Mini-4B-Instruct",
	hfFilePath: "nemotron-mini-4b-instruct-q5_k_m-imat-00001-of-00004.gguf",
	fileSizeInMegabytes: 3550,
	},
	"olmoe-1b-7b": {
	...createDefaultModelConfig(),
	label: "OLMoE 7B [1B]",
	hfRepoId: "Felladrin/gguf-sharded-Q3_K_XL-OLMoE-1B-7B-0924-Instruct",
	hfFilePath: "OLMoE-1B-7B-0924-Instruct-Q3_K_XL.shard-00001-of-00050.gguf",
	fileSizeInMegabytes: 3700,
	contextSize: 3584,
	},

For example, the files from the first model (hfRepoId: "Felladrin/gguf-Q8_0-SmolLM2-135M-Instruct") can be downloaded from https://huggingface.co/Felladrin/gguf-Q8_0-SmolLM2-135M-Instruct.