skillsafe-ai/flan-t5-small

FLAN-T5 small text-to-text (summarise, rewrite, Q&A)

Browser-ready import artifacts for text2text-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/Xenova/flan-t5-small/tree/311454e83bc784267fd7eef5940ee854144abbec
Upstream SHA-256 / commit	311454e83bc784267fd7eef5940ee854144abbec
Recipe	recipes/flan-t5-small.yaml — sha256 f230ca2f7f0078a0242a525bec811e06afbcaefbde7a742567544b38ebf88b25
Toolchain	Python 3.12.13, torch 2.10.0, onnx 1.23.0, onnxruntime 1.30.0 on Darwin 25.6.0 arm64
Converted	2026-09-22T21:43:57+00:00

Files

file	class	size	SHA-256
config.json	bundle	0.00 MB	a405fa336368f00e595e3a3f8f445e0318996f6d277008df11a577425a58e28d
generation_config.json	bundle	0.00 MB	6a6163360ca1826ba86f43d38af8d46d960b14a43090b619a5604e0b0bbf896c
onnx/decoder_model_merged.onnx	registry (fp32)	222.00 MB	0bae0554dd4d1deaf6401bb2168ff7213167ffb877f6d7de4610d33e201f8142
onnx/decoder_model_merged_quantized.onnx	registry (q8)	56.59 MB	73e2e942503221d7844715a8d824f68d8a2e878483f1c923849c177f0c441df1
onnx/encoder_model.onnx	registry (fp32, q8)	134.90 MB	07153ef2fd6eae4ccc29e579c66d3829ac40bbfc26fdc0b6a590cd3f55422b67
special_tokens_map.json	bundle	0.00 MB	5c87151ef0f72a99d1f766a4c418bd2a1f90aaa30a8e22fe5eca9641daebb64f
spiece.model	bundle	0.75 MB	d60acb128cf7b7f2536e8f38a5b18a05535c9e14c7a355904270e15b0945ea86
tokenizer.json	bundle	2.31 MB	d7af4599a1914d04aaf44839418757f40ccca4c78033edeba369484978378335
tokenizer_config.json	bundle	0.00 MB	8463612fde6595bf924e405f08192e1c58887e71332db2b9b287052dd3e180e1

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

Imported as published upstream (no conversion). Each file is pinned by SHA-256 to its source; every ONNX file passed onnx.checker and a CPU smoke run under onnxruntime with zero-filled inputs at the declared shapes:

file	inputs	outputs	ms
onnx/decoder_model_merged.onnx	encoder_attention_mask[1, 8], input_ids[1, 4], encoder_hidden_states[1, 8, 512], past_key_values.0.decoder.key[1, 6, 1, 64], past_key_values.0.decoder.value[1, 6, 1, 64], past_key_values.0.encoder.key[1, 6, 8, 64], past_key_values.0.encoder.value[1, 6, 8, 64], past_key_values.1.decoder.key[1, 6, 1, 64], past_key_values.1.decoder.value[1, 6, 1, 64], past_key_values.1.encoder.key[1, 6, 8, 64], past_key_values.1.encoder.value[1, 6, 8, 64], past_key_values.2.decoder.key[1, 6, 1, 64], past_key_values.2.decoder.value[1, 6, 1, 64], past_key_values.2.encoder.key[1, 6, 8, 64], past_key_values.2.encoder.value[1, 6, 8, 64], past_key_values.3.decoder.key[1, 6, 1, 64], past_key_values.3.decoder.value[1, 6, 1, 64], past_key_values.3.encoder.key[1, 6, 8, 64], past_key_values.3.encoder.value[1, 6, 8, 64], past_key_values.4.decoder.key[1, 6, 1, 64], past_key_values.4.decoder.value[1, 6, 1, 64], past_key_values.4.encoder.key[1, 6, 8, 64], past_key_values.4.encoder.value[1, 6, 8, 64], past_key_values.5.decoder.key[1, 6, 1, 64], past_key_values.5.decoder.value[1, 6, 1, 64], past_key_values.5.encoder.key[1, 6, 8, 64], past_key_values.5.encoder.value[1, 6, 8, 64], past_key_values.6.decoder.key[1, 6, 1, 64], past_key_values.6.decoder.value[1, 6, 1, 64], past_key_values.6.encoder.key[1, 6, 8, 64], past_key_values.6.encoder.value[1, 6, 8, 64], past_key_values.7.decoder.key[1, 6, 1, 64], past_key_values.7.decoder.value[1, 6, 1, 64], past_key_values.7.encoder.key[1, 6, 8, 64], past_key_values.7.encoder.value[1, 6, 8, 64], use_cache_branch[1]	logits[1, 4, 32128], present.0.decoder.key[1, 6, 4, 64], present.0.decoder.value[1, 6, 4, 64], present.0.encoder.key[1, 6, 8, 64], present.0.encoder.value[1, 6, 8, 64], present.1.decoder.key[1, 6, 4, 64], present.1.decoder.value[1, 6, 4, 64], present.1.encoder.key[1, 6, 8, 64], present.1.encoder.value[1, 6, 8, 64], present.2.decoder.key[1, 6, 4, 64], present.2.decoder.value[1, 6, 4, 64], present.2.encoder.key[1, 6, 8, 64], present.2.encoder.value[1, 6, 8, 64], present.3.decoder.key[1, 6, 4, 64], present.3.decoder.value[1, 6, 4, 64], present.3.encoder.key[1, 6, 8, 64], present.3.encoder.value[1, 6, 8, 64], present.4.decoder.key[1, 6, 4, 64], present.4.decoder.value[1, 6, 4, 64], present.4.encoder.key[1, 6, 8, 64], present.4.encoder.value[1, 6, 8, 64], present.5.decoder.key[1, 6, 4, 64], present.5.decoder.value[1, 6, 4, 64], present.5.encoder.key[1, 6, 8, 64], present.5.encoder.value[1, 6, 8, 64], present.6.decoder.key[1, 6, 4, 64], present.6.decoder.value[1, 6, 4, 64], present.6.encoder.key[1, 6, 8, 64], present.6.encoder.value[1, 6, 8, 64], present.7.decoder.key[1, 6, 4, 64], present.7.decoder.value[1, 6, 4, 64], present.7.encoder.key[1, 6, 8, 64], present.7.encoder.value[1, 6, 8, 64]	5.1
onnx/decoder_model_merged_quantized.onnx	encoder_attention_mask[1, 8], input_ids[1, 4], encoder_hidden_states[1, 8, 512], past_key_values.0.decoder.key[1, 6, 1, 64], past_key_values.0.decoder.value[1, 6, 1, 64], past_key_values.0.encoder.key[1, 6, 8, 64], past_key_values.0.encoder.value[1, 6, 8, 64], past_key_values.1.decoder.key[1, 6, 1, 64], past_key_values.1.decoder.value[1, 6, 1, 64], past_key_values.1.encoder.key[1, 6, 8, 64], past_key_values.1.encoder.value[1, 6, 8, 64], past_key_values.2.decoder.key[1, 6, 1, 64], past_key_values.2.decoder.value[1, 6, 1, 64], past_key_values.2.encoder.key[1, 6, 8, 64], past_key_values.2.encoder.value[1, 6, 8, 64], past_key_values.3.decoder.key[1, 6, 1, 64], past_key_values.3.decoder.value[1, 6, 1, 64], past_key_values.3.encoder.key[1, 6, 8, 64], past_key_values.3.encoder.value[1, 6, 8, 64], past_key_values.4.decoder.key[1, 6, 1, 64], past_key_values.4.decoder.value[1, 6, 1, 64], past_key_values.4.encoder.key[1, 6, 8, 64], past_key_values.4.encoder.value[1, 6, 8, 64], past_key_values.5.decoder.key[1, 6, 1, 64], past_key_values.5.decoder.value[1, 6, 1, 64], past_key_values.5.encoder.key[1, 6, 8, 64], past_key_values.5.encoder.value[1, 6, 8, 64], past_key_values.6.decoder.key[1, 6, 1, 64], past_key_values.6.decoder.value[1, 6, 1, 64], past_key_values.6.encoder.key[1, 6, 8, 64], past_key_values.6.encoder.value[1, 6, 8, 64], past_key_values.7.decoder.key[1, 6, 1, 64], past_key_values.7.decoder.value[1, 6, 1, 64], past_key_values.7.encoder.key[1, 6, 8, 64], past_key_values.7.encoder.value[1, 6, 8, 64], use_cache_branch[1]	logits[1, 4, 32128], present.0.decoder.key[1, 6, 4, 64], present.0.decoder.value[1, 6, 4, 64], present.0.encoder.key[1, 6, 8, 64], present.0.encoder.value[1, 6, 8, 64], present.1.decoder.key[1, 6, 4, 64], present.1.decoder.value[1, 6, 4, 64], present.1.encoder.key[1, 6, 8, 64], present.1.encoder.value[1, 6, 8, 64], present.2.decoder.key[1, 6, 4, 64], present.2.decoder.value[1, 6, 4, 64], present.2.encoder.key[1, 6, 8, 64], present.2.encoder.value[1, 6, 8, 64], present.3.decoder.key[1, 6, 4, 64], present.3.decoder.value[1, 6, 4, 64], present.3.encoder.key[1, 6, 8, 64], present.3.encoder.value[1, 6, 8, 64], present.4.decoder.key[1, 6, 4, 64], present.4.decoder.value[1, 6, 4, 64], present.4.encoder.key[1, 6, 8, 64], present.4.encoder.value[1, 6, 8, 64], present.5.decoder.key[1, 6, 4, 64], present.5.decoder.value[1, 6, 4, 64], present.5.encoder.key[1, 6, 8, 64], present.5.encoder.value[1, 6, 8, 64], present.6.decoder.key[1, 6, 4, 64], present.6.decoder.value[1, 6, 4, 64], present.6.encoder.key[1, 6, 8, 64], present.6.encoder.value[1, 6, 8, 64], present.7.decoder.key[1, 6, 4, 64], present.7.decoder.value[1, 6, 4, 64], present.7.encoder.key[1, 6, 8, 64], present.7.encoder.value[1, 6, 8, 64]	1.7
onnx/encoder_model.onnx	input_ids[1, 8], attention_mask[1, 8]	last_hidden_state[1, 8, 512]	1.6

Use in the browser

import * as ort from "onnxruntime-web";
const session = await ort.InferenceSession.create("https://huggingface.co/skillsafe-ai/flan-t5-small/resolve/main/onnx/decoder_model_merged.onnx", { executionProviders: ["webgpu", "wasm"] });

Contract (onnx/decoder_model_merged.onnx): input encoder_attention_mask int64 ['batch_size', 'encoder_sequence_length'], input_ids int64 ['batch_size', 'decoder_sequence_length'], encoder_hidden_states float32 ['batch_size', 'encoder_sequence_length', 512], past_key_values.0.decoder.key float32 ['batch_size', 6, 'past_decoder_sequence_length', 64], past_key_values.0.decoder.value float32 ['batch_size', 6, 'past_decoder_sequence_length', 64], past_key_values.0.encoder.key float32 ['batch_size', 6, 'encoder_sequence_length_out', 64], past_key_values.0.encoder.value float32 ['batch_size', 6, 'encoder_sequence_length_out', 64], past_key_values.1.decoder.key float32 ['batch_size', 6, 'past_decoder_sequence_length', 64], past_key_values.1.decoder.value float32 ['batch_size', 6, 'past_decoder_sequence_length', 64], past_key_values.1.encoder.key float32 ['batch_size', 6, 'encoder_sequence_length_out', 64], past_key_values.1.encoder.value float32 ['batch_size', 6, 'encoder_sequence_length_out', 64], past_key_values.2.decoder.key float32 ['batch_size', 6, 'past_decoder_sequence_length', 64], past_key_values.2.decoder.value float32 ['batch_size', 6, 'past_decoder_sequence_length', 64], past_key_values.2.encoder.key float32 ['batch_size', 6, 'encoder_sequence_length_out', 64], past_key_values.2.encoder.value float32 ['batch_size', 6, 'encoder_sequence_length_out', 64], past_key_values.3.decoder.key float32 ['batch_size', 6, 'past_decoder_sequence_length', 64], past_key_values.3.decoder.value float32 ['batch_size', 6, 'past_decoder_sequence_length', 64], past_key_values.3.encoder.key float32 ['batch_size', 6, 'encoder_sequence_length_out', 64], past_key_values.3.encoder.value float32 ['batch_size', 6, 'encoder_sequence_length_out', 64], past_key_values.4.decoder.key float32 ['batch_size', 6, 'past_decoder_sequence_length', 64], past_key_values.4.decoder.value float32 ['batch_size', 6, 'past_decoder_sequence_length', 64], past_key_values.4.encoder.key float32 ['batch_size', 6, 'encoder_sequence_length_out', 64], past_key_values.4.encoder.value float32 ['batch_size', 6, 'encoder_sequence_length_out', 64], past_key_values.5.decoder.key float32 ['batch_size', 6, 'past_decoder_sequence_length', 64], past_key_values.5.decoder.value float32 ['batch_size', 6, 'past_decoder_sequence_length', 64], past_key_values.5.encoder.key float32 ['batch_size', 6, 'encoder_sequence_length_out', 64], past_key_values.5.encoder.value float32 ['batch_size', 6, 'encoder_sequence_length_out', 64], past_key_values.6.decoder.key float32 ['batch_size', 6, 'past_decoder_sequence_length', 64], past_key_values.6.decoder.value float32 ['batch_size', 6, 'past_decoder_sequence_length', 64], past_key_values.6.encoder.key float32 ['batch_size', 6, 'encoder_sequence_length_out', 64], past_key_values.6.encoder.value float32 ['batch_size', 6, 'encoder_sequence_length_out', 64], past_key_values.7.decoder.key float32 ['batch_size', 6, 'past_decoder_sequence_length', 64], past_key_values.7.decoder.value float32 ['batch_size', 6, 'past_decoder_sequence_length', 64], past_key_values.7.encoder.key float32 ['batch_size', 6, 'encoder_sequence_length_out', 64], past_key_values.7.encoder.value float32 ['batch_size', 6, 'encoder_sequence_length_out', 64], use_cache_branch bool [1] → output logits float32 ['batch_size', 'decoder_sequence_length', 32128], present.0.decoder.key float32 ['batch_size', 6, 'past_decoder_sequence_length + 1', 64], present.0.decoder.value float32 ['batch_size', 6, 'past_decoder_sequence_length + 1', 64], present.0.encoder.key float32 ['batch_size', 6, 'encoder_sequence_length_out', 64], present.0.encoder.value float32 ['batch_size', 6, 'encoder_sequence_length_out', 64], present.1.decoder.key float32 ['batch_size', 6, 'past_decoder_sequence_length + 1', 64], present.1.decoder.value float32 ['batch_size', 6, 'past_decoder_sequence_length + 1', 64], present.1.encoder.key float32 ['batch_size', 6, 'encoder_sequence_length_out', 64], present.1.encoder.value float32 ['batch_size', 6, 'encoder_sequence_length_out', 64], present.2.decoder.key float32 ['batch_size', 6, 'past_decoder_sequence_length + 1', 64], present.2.decoder.value float32 ['batch_size', 6, 'past_decoder_sequence_length + 1', 64], present.2.encoder.key float32 ['batch_size', 6, 'encoder_sequence_length_out', 64], present.2.encoder.value float32 ['batch_size', 6, 'encoder_sequence_length_out', 64], present.3.decoder.key float32 ['batch_size', 6, 'past_decoder_sequence_length + 1', 64], present.3.decoder.value float32 ['batch_size', 6, 'past_decoder_sequence_length + 1', 64], present.3.encoder.key float32 ['batch_size', 6, 'encoder_sequence_length_out', 64], present.3.encoder.value float32 ['batch_size', 6, 'encoder_sequence_length_out', 64], present.4.decoder.key float32 ['batch_size', 6, 'past_decoder_sequence_length + 1', 64], present.4.decoder.value float32 ['batch_size', 6, 'past_decoder_sequence_length + 1', 64], present.4.encoder.key float32 ['batch_size', 6, 'encoder_sequence_length_out', 64], present.4.encoder.value float32 ['batch_size', 6, 'encoder_sequence_length_out', 64], present.5.decoder.key float32 ['batch_size', 6, 'past_decoder_sequence_length + 1', 64], present.5.decoder.value float32 ['batch_size', 6, 'past_decoder_sequence_length + 1', 64], present.5.encoder.key float32 ['batch_size', 6, 'encoder_sequence_length_out', 64], present.5.encoder.value float32 ['batch_size', 6, 'encoder_sequence_length_out', 64], present.6.decoder.key float32 ['batch_size', 6, 'past_decoder_sequence_length + 1', 64], present.6.decoder.value float32 ['batch_size', 6, 'past_decoder_sequence_length + 1', 64], present.6.encoder.key float32 ['batch_size', 6, 'encoder_sequence_length_out', 64], present.6.encoder.value float32 ['batch_size', 6, 'encoder_sequence_length_out', 64], present.7.decoder.key float32 ['batch_size', 6, 'past_decoder_sequence_length + 1', 64], present.7.decoder.value float32 ['batch_size', 6, 'past_decoder_sequence_length + 1', 64], present.7.encoder.key float32 ['batch_size', 6, 'encoder_sequence_length_out', 64], present.7.encoder.value float32 ['batch_size', 6, 'encoder_sequence_length_out', 64]. Opset 13.

Licence and attribution

FLAN-T5 small: Google, Apache License 2.0 (https://huggingface.co/google/flan-t5-small); ONNX export by Xenova (https://huggingface.co/Xenova/flan-t5-small).

Licence: Apache-2.0 — notice: https://huggingface.co/google/flan-t5-small/blob/main/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.