|
Download README.md from ggml-org-dev/tiny-Qwen3.5-9B-for-testing: direct link, hf CLI and curl.
- Browser
- Download file 1.5 kB
-
https://huggingface.co/ggml-org-dev/tiny-Qwen3.5-9B-for-testing/resolve/main/README.md
- Command line
-
hf download hf://ggml-org-dev/tiny-Qwen3.5-9B-for-testing/README.md
-
curl -L -o README.md https://huggingface.co/ggml-org-dev/tiny-Qwen3.5-9B-for-testing/resolve/main/README.md
1.5 kB
metadata
license: apache-2.0
base_model: Qwen/Qwen3.5-9B
tags:
- testing
tiny-Qwen3.5-9B-for-testing
This is a modified, damaged copy of Qwen/Qwen3.5-9B (revision
c202236), made only to test software. Its output is gibberish. Do not use it for anything else.
It keeps the architecture (hybrid linear attention / gated full attention, MTP layer, vision encoder), tokenizer and configs of Qwen3.5-9B, but it is small enough for CI. It is the base model of the tiny LoRA decision adapter ggml-org-dev/tiny-Bespoke-Nimble-9B-v3-for-testing: the llama.cpp converter downloads it and merges the adapter into it.
Changes from the original model
Sliced from the safetensors with slice_qwen35_lora.py, which follows the slice-model skill of llama.cpp:
- Text model: 4 of 32 layers (0, 1, 2, 3: 3 linear attention + 1 full attention), hidden size 4096 -> 32, FFN 12288 -> 128, heads 16 -> 2, KV heads 4 -> 1 (head_dim 256 kept), linear attention key heads 16 -> 2, value heads 32 -> 6 (head dims 128 kept)
- Token embedding and output: full vocabulary kept, narrowed to the hidden size
- MTP layer: kept, narrowed the same way
- Vision encoder: 27 -> 2 blocks, hidden size 1152 -> 144, heads 16 -> 2, FFN 4304 -> 288, output 4096 -> 32
License
Derived from Qwen3.5-9B by the Qwen team, under the Apache License 2.0.