Modelpublic

skillsafe-ai/smollm2-360m-instruct-q4f16

sourceHugging Faceapache-2.0updated 4d agoView on Hugging Face

0likes

Model Card

SmolLM2-360M-Instruct (q4f16_1, WebLLM)

Browser-ready mlc artifacts for text-generation, produced by SkillSafe's reproducible converter (`models/` in skillsafe.ai) from a pinned upstream source. Every byte here is derivable from that source plus the recipe below; nothing was edited by hand.

Provenance


Upstream	https://huggingface.co/mlc-ai/SmolLM2-360M-Instruct-q4f16_1-MLC/tree/3a622fd89e0216e8bb10c410c007c786baa8a033
Upstream SHA-256 / commit	`3a622fd89e0216e8bb10c410c007c786baa8a033`
Recipe	`recipes/smollm2-360m-instruct-q4f16.yaml` — sha256 `915e81797ae2b2e84233315a9aba031551ed631ab3ce95ed52767130812741c9`
Toolchain	Python 3.12.13, torch 2.10.0, onnx 1.23.0, onnxruntime 1.30.0 on Darwin 25.6.0 arm64
Converted	2026-09-22T19:01:10+00:00

Files

file	class	size	SHA-256
`lib/SmolLM2-360M-Instruct-q4f16_1-ctx4k_cs1k-webgpu.wasm`	registry-shared	5.62 MB	`3db2492c6b6c2ad5fa8b3969deaa6d6a345c1b05b3bd5f3d8a82b52de9d14ceb`
`merges.txt`	bundle	0.44 MB	`0b54e8aa4e53d5383e2e4bc635a56b43f9647f7b13832d5d9ecd8f82dac4f510`
`mlc-chat-config.json`	bundle	0.00 MB	`b7329699d6fdbf5a4f68a0c42b14250f4beab2cabbfa120fae6c924e7dcd8f2c`
`ndarray-cache.json`	bundle	0.12 MB	`55a84311c4516c04218947b0633a5bb9365f9873ecdb69757f5e308ef87c515b`
`params_shard_0.bin`	registry	31.91 MB	`5d5807a95573ed1d72e8c7f2bac2911ba1ad78bb486ab9745693eb51e8b7616c`
`params_shard_1.bin`	registry	31.66 MB	`7f6901def1d79db3c3eabca3e271ce7af131d3bec599971b98ad66f3cdbc3729`
`params_shard_2.bin`	registry	31.66 MB	`bb2d45b7b35602e0667575de94f750dd1ec5cd5d01a840c2f2fb9bdc553fc66a`
`params_shard_3.bin`	registry	31.66 MB	`0f5c419205d2d03508dbd10f937c1508902ae4ea5330a6493b3e2e7ef0df7ed0`
`params_shard_4.bin`	registry	31.66 MB	`959bd10c37b9634a077e636be15e59059bea6a9a152de2e7bab81d248d0644e6`
`params_shard_5.bin`	registry	31.66 MB	`99c3b1c9ef35f322013e07c607ae9c271e5895138e347a7e60acbbe1645407e8`
`params_shard_6.bin`	registry	3.96 MB	`09b74f2ad1f35fcacc6567b6b6b0fc4a69dd6191dd94fad413a252096693c8c2`
`tensor-cache.json`	bundle	0.12 MB	`55a84311c4516c04218947b0633a5bb9365f9873ecdb69757f5e308ef87c515b`
`tokenizer.json`	bundle	2.01 MB	`9ca9acddb6525a194ec8ac7a87f24fbba7232a9a15ffa1af0c1224fcd888e47c`
`tokenizer_config.json`	bundle	0.00 MB	`a27f638bd2831f5c3dea654a75838930f2b11fbe550c4d4e1d5d7bd07157b2ee`
`vocab.json`	bundle	0.76 MB	`82b84012e3add4d01d12ba14442026e49b8cbbaead1f79ecf3d919784f82dc79`

registry files are parameter files served from models.skillsafe.ai once vetted; bundle files ship inside an app; registry-shared is a runtime library reused by every model of the same architecture.

Verification

Shard sizes and md5s match ndarray-cache.json; the implied parameter count matches the checkpoint (361,821,120 parameters, 4.50 bits/param); the parameter table embedded in the WebGPU model library matches all 323 tensors of these weights.

Use in the browser (WebLLM)

import { CreateMLCEngine } from "@mlc-ai/web-llm";
const engine = await CreateMLCEngine("smollm2-360m-instruct-q4f16", {
  appConfig: { model_list: [{
    model: "https://huggingface.co/skillsafe-ai/smollm2-360m-instruct-q4f16",
    model_id: "smollm2-360m-instruct-q4f16",
    model_lib: "https://huggingface.co/skillsafe-ai/smollm2-360m-instruct-q4f16/resolve/main/lib/SmolLM2-360M-Instruct-q4f16_1-ctx4k_cs1k-webgpu.wasm",
    overrides: { context_window_size: 4096, prefill_chunk_size: 1024 },
  }] },
});

The model library is compiled for context window 4096 / prefill chunk 1024; mlc-chat-config.json says 8192 / 8192, so set the overrides above.

Licence and attribution

SmolLM2-360M-Instruct: Hugging Face TB, Apache License 2.0. https://huggingface.co/HuggingFaceTB/SmolLM2-360M-Instruct — q4f161 weights published by MLC (https://huggingface.co/mlc-ai/SmolLM2-360M-Instruct-q4f161-MLC).

Licence: Apache-2.0 — notice: https://huggingface.co/HuggingFaceTB/SmolLM2-360M-Instruct/blob/a10cc1512eabd3dde888204e902eca88bddb4951/README.md. The conversion recipe and this model card are part of the SkillSafe repository and carry its licence; the weights remain under the upstream licence above.

The full manifest.json in this repo records the recipe, sources, toolchain (including the uv.lock hash) and per-file verification numbers.