Downloads · 30 days
0
smollm2-135m-onnx-patched
Vibelet/smollm2-135m-onnx-patched
smollm2-135m-onnx-patched is a machine learning model from Vibelet. Use it for the machine learning task on the model card, and read the license before you ship it in a product.
Patched derivative of onnx-community/SmolLM2-135M-ONNX (onnx/modelint8.onnx), produced by appending the residual-stream tensors as graph outputs so a single forward pass yields all 31 hidden states:
Access
Public
Updated Aug 17, 2026
Repo size
136 MB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.onnx136 MB · 100%
At a glance
- Access
- Public
- Created
- Aug 17, 2026
- Updated
- Aug 17, 2026
- SHA
- fcc21559
Model card
From the Hugging Face model README
SmolLM2-135M ONNX (patched hidden states)
Patched derivative of onnx-community/SmolLM2-135M-ONNX
(onnx/model_int8.onnx), produced by appending the residual-stream tensors
as graph outputs so a single forward pass yields all 31 hidden states:
/model/layers.{0..29}/input_layernorm/Cast_output_0(embeddings, then the residual stream after each of the first 29 layers)/model/norm/Cast_output_0(residual after layer 29)
Each is fp32 [1, seq, 576]. The stock export exposes only logits + KV caches,
so this patch exists to feed a browser-side per-layer projection cabinet.
The weight tensors are unchanged.
Hosted here so that a small static site can fetch the model without charging its hosting provider 130 MB per visitor.
- Base model: HuggingFaceTB/SmolLM2-135M
- ONNX export:
onnx-community/SmolLM2-135M-ONNX - License: Apache-2.0
- Created
- Aug 17, 2026
- Updated
- Aug 17, 2026