Downloads · 30 days
6
8% of all-time downloads
drodriguezhdez/qwen35-in_proj_qkv-repro
qwen35-in_proj_qkv-repro is a machine learning model from drodriguezhdez. Use it for the machine learning task on the model card, and read the license before you ship it in a product. It is set up for peft. The card lists the license as apache-2.0.
A tiny, synthetic, random-weight LoRA adapter whose only purpose is to reproduce a bug in SGLang's LoRA support for Qwen/Qwen3.6-35B-A3B (arch Qwen35MoeForConditionalGeneration) — reported as sgl-project/sglang30168.…
Downloads · 30 days
6
8% of all-time downloads
All-time downloads
72
Public
Repo size
3.9 MB
Likes
0
Public
Click a slice to open those files.
.safetensors3.9 MB · 100%
From the Hugging Face model README
A tiny, synthetic, random-weight LoRA adapter whose only purpose is to reproduce a bug in
SGLang's LoRA support for Qwen/Qwen3.6-35B-A3B
(arch Qwen3_5MoeForConditionalGeneration) — reported as
sgl-project/sglang#30168. This is not a trained
model — do not use it for inference quality; the weights are random.
Serving the base model with --enable-lora and this adapter crashes SGLang (v0.5.14) at startup,
during LoRA memory-pool initialization:
NotImplementedError: get_hidden_dim not implemented for in_proj_qkv
NotImplementedError: get_hidden_dim not implemented for in_proj_z
HF transformers exposes the Gated-DeltaNet input projection as a fused in_proj_qkv (q, k, v in one
matmul) plus a separate in_proj_z, so PEFT emits LoRA weights under those 2-way names. SGLang's
Qwen3.5/GDN LoRA path only knows the fully-fused in_proj_qkvz, so get_hidden_dim has no case for the
native names and the server aborts.
python3 -m sglang.launch_server --model-path Qwen/Qwen3.6-35B-A3B \
--enable-lora --lora-paths repro=drodriguezhdez/qwen35-in_proj_qkv-repro \
--max-lora-rank 8 --lora-backend triton
target_modules: ["in_proj_qkv", "in_proj_z"], rank 4, fp16, across the 30 Gated-DeltaNet layers.in_proj_qkv.lora_A [4, 2048], in_proj_qkv.lora_B [8192, 4] (fused q,k,v = 2048+2048+4096),
in_proj_z.lora_A [4, 2048], in_proj_z.lora_B [4096, 4].These names and shapes are identical to those emitted by a real PEFT adapter trained on this base model.
The crash depends only on the adapter's module names and shapes, never the values — so this adapter carries randomly-initialized weights (no proprietary data) and reproduces the crash identically to a real trained adapter with the same targets.
The fully-fused in_proj_qkvz layout is served correctly; repacking the adapter offline into it (split
in_proj_qkv's output back into q/k/v, set target_modules: ["in_proj_qkvz"]) lets it serve and apply
correctly on v0.5.14.