Add files using upload-large-folder tool

.gitattributes

@@ -33,3 +33,8 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+Qwen3.5-4B-Q3_K_S.gguf filter=lfs diff=lfs merge=lfs -text
+Qwen3.5-4B-Q4_K_M.gguf filter=lfs diff=lfs merge=lfs -text
+Qwen3.5-4B-Q5_K_M.gguf filter=lfs diff=lfs merge=lfs -text
+Qwen3.5-4B-Q6_K.gguf filter=lfs diff=lfs merge=lfs -text
+Qwen3.5-4B-Q8_0.gguf filter=lfs diff=lfs merge=lfs -text

Qwen3.5-4B-Q3_K_S.gguf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d5c3667c3eb22674b5d89184f5903784f338ff3fe94f19eb8c19249aa9f6149f
+size 1980940384

Qwen3.5-4B-Q4_K_M.gguf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:036473af3820990fc56ca6bff29d267acf3d0d0726ed1cfda8f2510ce683bdb3
+size 2606333024

Qwen3.5-4B-Q5_K_M.gguf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ee3f496e4ad5f814409c0aa4eb1841e59106b666ba03ec7ae6816d2f50563d14
+size 2706322784

Qwen3.5-4B-Q6_K.gguf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:57213a5168c922a21f6752a8c85593daf08c3618dfccaeae3560d08410266f77
+size 3251025504

Qwen3.5-4B-Q8_0.gguf

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6716f24df57ead0245b3a0582584934799dbff532d6959d37671b0d5293b5eb8
+size 4482403424

README.md

@@ -0,0 +1,95 @@
+---
+language: [en, zh]
+license: apache-2.0
+library_name: gguf
+base_model: Qwen/Qwen3.5-4B
+tags: [quantized, gguf, mixed-precision, shapelearn, shapelearn3, qwen3]
+pipeline_tag: text-generation
+---
+
+# Qwen3.5-4B — ShapeLearn3 Mixed-Precision GGUF
+
+Mixed-precision GGUF quantizations of [Qwen3.5-4B](https://huggingface.co/Qwen/Qwen3.5-4B)
+— the first ShapeLearn release of this model. ShapeLearn3 keeps the Hessian-sensitivity
+front-end to set each level's bit budget, but hands the per-tensor allocation to an
+**error-minimizing solver** (built on llama.cpp's `--target-bpw`) that distributes bits
+across *every* tensor — including the hybrid DeltaNet/SSM tensors — to minimize
+imatrix-weighted quantization error at the target size.
+
+## Available Quantizations
+
+KLD vs the BF16 source is the primary quality metric (mean and the robust 99.9th
+percentile); wikitext-2 perplexity is reported alongside.
+
+| File | BPW | Size | wiki PPL ↓ | KL-mean ↓ | KL-99.9% ↓ | Use Case |
+|---|---|---|---|---|---|---|
+| [`Qwen3.5-4B-Q8_0.gguf`](./Qwen3.5-4B-Q8_0.gguf) | 8.5 | 4.48 GB | 8.638 | 0.0028 | 0.073 | Near-lossless reference |
+| [`Qwen3.5-4B-Q6_K.gguf`](./Qwen3.5-4B-Q6_K.gguf) | 6.2 | 3.25 GB | 8.767 | 0.0125 | 0.396 | High quality |
+| [`Qwen3.5-4B-Q5_K_M.gguf`](./Qwen3.5-4B-Q5_K_M.gguf) | 5.1 | 2.71 GB | 9.176 | 0.0255 | 0.803 | Balanced quality and size |
+| [`Qwen3.5-4B-Q4_K_M.gguf`](./Qwen3.5-4B-Q4_K_M.gguf) | 5.0 | 2.61 GB | 9.246 | 0.0292 | 0.952 | Best quality-to-size ratio |
+| [`Qwen3.5-4B-Q3_K_S.gguf`](./Qwen3.5-4B-Q3_K_S.gguf) | 3.8 | 1.98 GB | 9.358 | 0.0978 | 3.583 | Maximum compression |
+
+**Recommended:** Q4_K_M — KL-mean 0.029 at 2.61 GB.
+
+## How It Compares
+
+Same harness, same metrics, against a fixed per-suffix LP-recipe baseline at matched BPW
+(within 0.5%) — the head-to-head that motivated ShapeLearn3's allocator:
+
+| Level | LP recipe KL-mean | **ShapeLearn3** KL-mean |
+|---|---|---|
+| Q3_K_S | 0.1525 | **0.0978** |
+| Q4_K_M | 0.0352 | **0.0292** |
+| Q5_K_M | 0.0258 | **0.0255** |
+
+The allocator wins KL divergence at every level — largest at aggressive quantization
+(Q3_K_S: −36% KL-mean) — and our internal worst-token measurements improve at every level
+as well. A key reason: it allocates bits to the hybrid DeltaNet tensors (`attn_qkv`,
+`attn_gate`, `ssm_*`) that a standard 7-suffix recipe leaves at the base type — and our
+tensor-health scan shows those exact tensors are the statistical outliers of this
+architecture.
+
+## Key Sensitivity Findings (Qwen3.5-4B)
+
+- **blk.3 (early layer) is most sensitive** — the same early-layer pattern as Qwen3.5-9B,
+  and the opposite of dense Qwen3-4B-Instruct (blk.34). The hybrid Qwen3.5 family
+  concentrates sensitivity early.
+- Attention **K projections are consistently ≥ V** in sensitivity.
+- **DeltaNet/SSM tensors are distribution outliers** (high kurtosis `ssm_conv1d`, shifted
+  `ssm_alpha/beta/out`, `attn_qkv`, `attn_gate` vs same-role peers) — covering them in the
+  allocation matters; `ssm_conv1d` itself is kept at F32 by llama.cpp.
+
+## How It Works
+
+1. **Hessian sensitivity** — compute H_diag = mean(X²) per layer on calibration data; this
+   sets each level's overall bit budget.
+2. **Error-minimizing per-tensor allocation** — an imatrix-weighted solver (llama.cpp
+   `--target-bpw`) assigns a quant type to every tensor to minimize total quantization
+   error at the target BPW, covering attention, FFN, and the hybrid DeltaNet/SSM tensors.
+3. **imatrix** — importance matrix computed over wikitext guides the per-tensor error.
+4. **GGUF export** — produced with stock `llama-quantize`.
+
+## Usage
+
+```bash
+hf download sh111111111111111/Qwen3.5-4B-ShapeLearn3-GGUF \
+    Qwen3.5-4B-Q4_K_M.gguf --local-dir .
+
+llama-cli    -m Qwen3.5-4B-Q4_K_M.gguf -cnv
+llama-server -m Qwen3.5-4B-Q4_K_M.gguf --port 8080
+```
+
+> Note: Qwen3.5 GGUFs are not currently runnable in Ollama (vision/mmproj handling is not
+> yet supported there); use llama.cpp or LM Studio.
+
+## Benchmark Details
+
+NVIDIA GB10 ATOM (128 GB unified memory, aarch64). llama.cpp with `--target-bpw`
+(PR #15550). KLD via `llama-perplexity --kl-divergence` against BF16-source logits over
+wikitext-2 (mean / median / 99.9th percentile reported; the single-token KL-max is
+omitted as an unstable order statistic). wikitext-2 PPL via `llama-perplexity -c 2048`.
+Downstream (HellaSwag / WinoGrande / ARC / MMLU) tracked internally.
+
+## License
+
+Apache 2.0, inherited from [Qwen3.5-4B](https://huggingface.co/Qwen/Qwen3.5-4B).