---

license: mit

base_model:

- deepseek-ai/DeepSeek-V4-Flash-Vision-Exp

library_name: llama.cpp

tags:

- gguf

- layer-library

- multi-stage

- ik_llama.cpp

- dreamcatcher

- xautonomics

---

DeepSeek-V4-Flash-Vision-Exp-UD-Q4_K_XL.LAYR.GGUF

A per-layer GGUF library of Unsloth's unsloth/DeepSeek-V4-Flash-Vision-Exp-GGUF (UD-Q4_K_XL). These are the same quantized weights, sliced once into one file per transformer block plus the shared parts, so any contiguous layer window A, B) can be assembled at load time. It's built for [dreamcatcher, xAutonomics' fork of ik_llama.cpp, which serves a library either in a single process or split across hosts.

| | |

|---|---|

| base model | deepseek-ai/DeepSeek-V4-Flash-Vision-Exp |

| source quant | unsloth/DeepSeek-V4-Flash-Vision-Exp-GGUF · UD-Q4_K_XL · pinned revision 37044a3cff · 5 shards |

| architecture | deepseek4 · 43 blocks |

| files | 46 (embd, layer, other, output) · 144.7 GiB |

| integrity | manifest.json: blake2b-128 per whole file and per tensor |

| sliced with | layer_distribution.slice (the slicer commit is in the SLICER_COMMIT file); nothing was re-quantized |

Support status

Measured on the dreamcatcher tree (status as of c2f8b680; the linked document is authoritative):

deepseek4 — CUDA verified. Monolith verified on CUDA (UD-Q4_K_XL, revision

37044a3c, tensor-identical to the measured copy; 1328/1328 tensor hashes, full

43-layer forward pass byte-identical). Library windows fixed (meta#92): ring

token-identical 64/64 on CUDA with the meta#92 fix; llama-server --model-dir

is available (meta#90; not yet measured on this architecture). Vulkan:

degenerate on NVIDIA Vulkan (coopmat1); RDNA3/ANV not measured; not quant-related

(meta#96).

Only model files from huggingface.co/xautonomics are supported by dreamcatcher. Other GGUFs, including libraries you slice yourself, may load but are unsupported. Do not rename the files: the loader expects blk-NNNNN.gguf / parts-*.gguf / manifest.json exactly.

Layout

blk-00000.gguf … blk-00042.gguf   one transformer block each
parts-embd.gguf                          token embedding (every stage loads it)
parts-output.gguf                        output norm + head (every stage loads it)
parts-other.gguf                         remaining non-block tensors (every stage loads it)
manifest.json                            file list, sizes, per-file + per-tensor hashes, source provenance

A stage that serves layers [A, B) needs blk-A … blk-(B-1) plus the parts-* files and manifest.json.

Use

Single process (the whole model on one host):

llama-server --model-dir /models/DeepSeek-V4-Flash-Vision-Exp-UD-Q4_K_XL.LAYR.GGUF

Multi-host ring (stages exchange hidden states over TCP; each host needs low-latency, direct node-to-node links — see the dreamcatcher README):

llama-stage-runner --role tail --listen 8081 --model-dir /models/DeepSeek-V4-Flash-Vision-Exp-UD-Q4_K_XL.LAYR.GGUF --layers 24,43
llama-stage-runner --role head --connect <tail-host>:8081 --model-dir /models/DeepSeek-V4-Flash-Vision-Exp-UD-Q4_K_XL.LAYR.GGUF --layers 0,24

Plan a split for your hosts (weights-only screen; the proof is a survived prefill), and verify the download before launching:

python3 -m layer_distribution.plan /models/DeepSeek-V4-Flash-Vision-Exp-UD-Q4_K_XL.LAYR.GGUF/manifest.json --node a:16GiB:200GiB:8080 --node b:24GiB:200GiB:8081
python3 -c "from layer_distribution import verify; r = verify('/models/DeepSeek-V4-Flash-Vision-Exp-UD-Q4_K_XL.LAYR.GGUF', '/models/DeepSeek-V4-Flash-Vision-Exp-UD-Q4_K_XL.LAYR.GGUF/manifest.json'); print(r.passed, r.hash_verified)"

Provenance and license

Weights are byte-for-byte the tensors of unsloth/DeepSeek-V4-Flash-Vision-Exp-GGUF at revision 37044a3cff5ed45e3832f3e833bf1cff91f7f168 (Unsloth's dynamic quantization of deepseek-ai/DeepSeek-V4-Flash-Vision-Exp). Source shard names and sizes are recorded in manifest.json (source.shards[], source.content_hash). License follows the base model: mit, see LICENSE. Credit to the base-model authors and to Unsloth for the quantization.