cstr
/

deepseek-ocr2-crispembed-GGUF

Mixture of Experts

Model card Files Files and versions

deepseek-ocr2-crispembed-GGUF

Commit History

docs: add provenance / EU AI Act Art. 53 note

731d937
verified

cstr commited on 24 days ago

f16 stacked MoE experts (#4) — canonical source for requant

ec0fda0
verified

cstr commited on Jul 13

q4_k stacked MoE experts (#4) — ffn_*_exps 3D tensors, byte-identical to per-expert

f8d16c5
verified

cstr commited on Jul 13

Q4_K (reconverted, fixed tokenizer)

6934d47
verified

cstr commited on Jul 2

F16 (reconverted, fixed tokenizer)

a798d6a
verified

cstr commited on Jul 2

Upload deepseek-ocr2-f16.gguf with huggingface_hub

fa590b6
verified

cstr commited on Jun 20

Upload README.md with huggingface_hub

a23a066
verified

cstr commited on Jun 20

F16 (reconverted, fixed tokenizer)

bc7dd3c
verified

cstr commited on Jun 19

Q4_K (reconverted, fixed tokenizer)

a465ab6
verified

cstr commited on Jun 19

Q8_0 (reconverted, fixed tokenizer)

3b7b4c0
verified

cstr commited on Jun 19

Add parity reference (sam+qwen2enc+projector, 64x64 test)

1ed1ebb
verified

cstr commited on Jun 16

Add DeepSeek-OCR-2 F16 GGUF (3.4B MoE, Apache-2.0)

306ae63
verified

cstr commited on Jun 16

Add README with architecture details

c89604f
verified

cstr commited on Jun 16

initial commit

54ae2ab
verified

cstr commited on Jun 16