mirror of
https://github.com/mudler/LocalAI.git
synced 2026-11-06 04:23:15 -05:00
feat(gallery): add MiniCPM-OCR 1.0 models
Add OpenBMB's 1.3B Chinese/English document parser to the gallery as
three llama.cpp entries: the recommended Q8_0 build, a smaller Q4_K_M
build for constrained hosts, and an optional DSpark build that pairs the
Q8_0 target with the official speculative draft.
All entries install the required F16 4x vision projector, default to an
8,192-token single-slot context, and keep thinking disabled so
transcriptions land in the response content. Q8_0 is the declaring entry;
the other two are selectable variants.
Tested text, table, and mixed Chinese/English page OCR through LocalAI
on an RTX 4090; the DSpark build reached about 0.96 draft acceptance. The
model's projector type is new in llama.cpp, so the entries need a backend
built from the current master pin; the latest release backend predates it.
Assisted-by: claude code:opus5.5
Signed-off-by: Caldalis <zyk_computer@163.com>
(cherry picked from commit 91ff3f06cc)
This commit is contained in:
Caldalis authored and Ettore Di Giacinto committed
1 parent
c82ae6e953
commit
f88fa99e1d
3 files changed
+473
No files matched your search
@@ -0,0 +1,125 @@
|
||||
// SPDX-License-Identifier: MIT
|
||||
package gallery_test
|
||||
|
||||
import (
|
||||
. "github.com/onsi/ginkgo/v2"
|
||||
. "github.com/onsi/gomega"
|
||||
|
||||
"github.com/mudler/LocalAI/core/gallery"
|
||||
)
|
||||
|
||||
// MiniCPM-OCR only runs as a multimodal model: the 4x projector is required,
|
||||
// and the entry must keep thinking disabled so the transcription lands in the
|
||||
// response content instead of the reasoning channel. The DSpark build adds a
|
||||
// second model file that must be named as the draft, not as a variant's
|
||||
// primary weights.
|
||||
var _ = Describe("gallery/index.yaml MiniCPM-OCR entries", func() {
|
||||
const (
|
||||
modelDir = "llama-cpp/models/minicpm-ocr-1.0/"
|
||||
mmproj = modelDir + "mmproj-model-f16-4x.gguf"
|
||||
mmprojSHA = "ceed82261b3b1380c3165f8b37c10117a50093be1362cbfeeb692fa036444682"
|
||||
q8 = modelDir + "MiniCPM-OCR-1_0-Q8_0.gguf"
|
||||
q8SHA = "791486a476bdacb48acb3f31aa3a18dbb5be87e723a4074bcd13a37d61c852e2"
|
||||
q4 = modelDir + "MiniCPM-OCR-1_0-Q4_K_M.gguf"
|
||||
q4SHA = "c80a703400633f298cc2fddf6d493a541be36918c0801c0b1bf3f14c5703ce49"
|
||||
dspark = modelDir + "dspark-MiniCPM-OCR-1_0-Q8_0.gguf"
|
||||
dsparkSHA = "fd8f64bb62cf5a5c6b4c0999bfa58f64453690c9f75f1124ab613dfb44a822da"
|
||||
ocrBase = "https://huggingface.co/openbmb/MiniCPM-OCR-1.0-gguf/resolve/9ba8c0db38ec333bdd83d44a88d25c6f1d748ba4/"
|
||||
dsparkURL = "https://huggingface.co/openbmb/MiniCPM-OCR-1.0-DSpark-gguf/resolve/265c4dc398c8d5c0ac56a4dd155a5a97b2037226/"
|
||||
)
|
||||
|
||||
byName := func() map[string]gallery.GalleryModel {
|
||||
entries, err := loadGalleryIndex()
|
||||
Expect(err).ToNot(HaveOccurred())
|
||||
return indexEntriesByName(entries)
|
||||
}
|
||||
fileByName := func(e gallery.GalleryModel, name string) (gallery.File, bool) {
|
||||
for _, f := range e.AdditionalFiles {
|
||||
if f.Filename == name {
|
||||
return f, true
|
||||
}
|
||||
}
|
||||
return gallery.File{}, false
|
||||
}
|
||||
|
||||
It("declares the Q8, Q4_K_M and DSpark builds as one selectable family", func() {
|
||||
entries := byName()
|
||||
base, ok := entries["minicpm-ocr-1.0-q8"]
|
||||
Expect(ok).To(BeTrue())
|
||||
|
||||
variants := make([]string, 0, len(base.Variants))
|
||||
for _, v := range base.Variants {
|
||||
variants = append(variants, v.Model)
|
||||
}
|
||||
Expect(variants).To(ConsistOf("minicpm-ocr-1.0-q4km", "minicpm-ocr-1.0-dspark"))
|
||||
|
||||
for _, name := range []string{"minicpm-ocr-1.0-q8", "minicpm-ocr-1.0-q4km", "minicpm-ocr-1.0-dspark"} {
|
||||
e, ok := entries[name]
|
||||
Expect(ok).To(BeTrue(), name)
|
||||
Expect(e.License).To(Equal("apache-2.0"), name)
|
||||
Expect(e.Overrides).To(HaveKeyWithValue("backend", "llama-cpp"), name)
|
||||
Expect(e.Overrides).To(HaveKeyWithValue("context_size", 8192), name)
|
||||
Expect(e.Overrides).To(HaveKeyWithValue("mmproj", mmproj), name)
|
||||
Expect(e.Overrides["known_usecases"]).To(ConsistOf("chat", "vision"), name)
|
||||
|
||||
tmpl, ok := e.Overrides["template"].(map[string]any)
|
||||
Expect(ok).To(BeTrue(), name)
|
||||
Expect(tmpl).To(HaveKeyWithValue("use_tokenizer_template", true), name)
|
||||
}
|
||||
})
|
||||
|
||||
It("keeps thinking disabled and selects the expected weights", func() {
|
||||
entries := byName()
|
||||
|
||||
q8Entry := entries["minicpm-ocr-1.0-q8"]
|
||||
Expect(q8Entry.Overrides["options"]).To(ConsistOf("use_jinja:true", "enable_reasoning:0", "parallel:1"))
|
||||
Expect(q8Entry.Overrides["parameters"]).To(HaveKeyWithValue("model", q8))
|
||||
|
||||
q4Entry := entries["minicpm-ocr-1.0-q4km"]
|
||||
Expect(q4Entry.Overrides["options"]).To(ConsistOf("use_jinja:true", "enable_reasoning:0", "parallel:1"))
|
||||
Expect(q4Entry.Overrides["parameters"]).To(HaveKeyWithValue("model", q4))
|
||||
|
||||
dsparkEntry := entries["minicpm-ocr-1.0-dspark"]
|
||||
Expect(dsparkEntry.Overrides["options"]).To(ConsistOf(
|
||||
"use_jinja:true", "enable_reasoning:0", "parallel:1", "spec_type:draft-dspark", "spec_n_max:15"))
|
||||
Expect(dsparkEntry.Overrides).To(HaveKeyWithValue("draft_model", dspark))
|
||||
Expect(dsparkEntry.Overrides["parameters"]).To(HaveKeyWithValue("model", q8))
|
||||
})
|
||||
|
||||
It("pins every weight and projector to the published sha256", func() {
|
||||
entries := byName()
|
||||
want := map[string]map[string]string{
|
||||
"minicpm-ocr-1.0-q8": {
|
||||
q8: q8SHA,
|
||||
mmproj: mmprojSHA,
|
||||
},
|
||||
"minicpm-ocr-1.0-q4km": {
|
||||
q4: q4SHA,
|
||||
mmproj: mmprojSHA,
|
||||
},
|
||||
"minicpm-ocr-1.0-dspark": {
|
||||
q8: q8SHA,
|
||||
dspark: dsparkSHA,
|
||||
mmproj: mmprojSHA,
|
||||
},
|
||||
}
|
||||
for name, files := range want {
|
||||
e := entries[name]
|
||||
Expect(e.AdditionalFiles).To(HaveLen(len(files)), name)
|
||||
for filename, sha := range files {
|
||||
f, ok := fileByName(e, filename)
|
||||
Expect(ok).To(BeTrue(), name+": "+filename)
|
||||
Expect(f.SHA256).To(Equal(sha), name+": "+filename)
|
||||
}
|
||||
}
|
||||
|
||||
q8File, _ := fileByName(entries["minicpm-ocr-1.0-q8"], q8)
|
||||
Expect(q8File.URI).To(Equal(ocrBase + "MiniCPM-OCR-1_0-Q8_0.gguf"))
|
||||
mmprojFile, _ := fileByName(entries["minicpm-ocr-1.0-q8"], mmproj)
|
||||
Expect(mmprojFile.URI).To(Equal(ocrBase + "mmproj-model-f16-4x.gguf"))
|
||||
q4File, _ := fileByName(entries["minicpm-ocr-1.0-q4km"], q4)
|
||||
Expect(q4File.URI).To(Equal(ocrBase + "MiniCPM-OCR-1_0-Q4_K_M.gguf"))
|
||||
dsparkFile, _ := fileByName(entries["minicpm-ocr-1.0-dspark"], dspark)
|
||||
Expect(dsparkFile.URI).To(Equal(dsparkURL + "dspark-MiniCPM-OCR-1_0-Q8_0.gguf"))
|
||||
})
|
||||
})
|
||||
@@ -31223,6 +31223,168 @@
|
||||
model: OvisOCR2-Q8_0.gguf
|
||||
template:
|
||||
use_tokenizer_template: true
|
||||
- url: github:mudler/LocalAI/gallery/virtual.yaml@master
|
||||
name: minicpm-ocr-1.0-q8
|
||||
description: |
|
||||
MiniCPM-OCR 1.0 is OpenBMB's 1.3B document-parsing model for Chinese and
|
||||
English text, tables, formulas, charts, and seals. This entry uses the
|
||||
official Q8_0 GGUF weights and the required F16 4x vision projector
|
||||
through llama.cpp, with an 8,192-token default context. Prompt the model
|
||||
with "Text Recognition:" for text, "Table Recognition:" for tables, or
|
||||
"Formula Recognition:" for formulas. Table output is the model's OTSL
|
||||
markup, which clients can convert to HTML.
|
||||
license: apache-2.0
|
||||
urls:
|
||||
- https://huggingface.co/openbmb/MiniCPM-OCR-1.0
|
||||
- https://huggingface.co/openbmb/MiniCPM-OCR-1.0-gguf
|
||||
icon: https://avatars.githubusercontent.com/u/89920203
|
||||
tags:
|
||||
- llm
|
||||
- gguf
|
||||
- gpu
|
||||
- cpu
|
||||
- image-to-text
|
||||
- ocr
|
||||
- document-parsing
|
||||
- multimodal
|
||||
- vision
|
||||
- minicpm
|
||||
files:
|
||||
- filename: llama-cpp/models/minicpm-ocr-1.0/MiniCPM-OCR-1_0-Q8_0.gguf
|
||||
sha256: 791486a476bdacb48acb3f31aa3a18dbb5be87e723a4074bcd13a37d61c852e2
|
||||
uri: https://huggingface.co/openbmb/MiniCPM-OCR-1.0-gguf/resolve/9ba8c0db38ec333bdd83d44a88d25c6f1d748ba4/MiniCPM-OCR-1_0-Q8_0.gguf
|
||||
- filename: llama-cpp/models/minicpm-ocr-1.0/mmproj-model-f16-4x.gguf
|
||||
sha256: ceed82261b3b1380c3165f8b37c10117a50093be1362cbfeeb692fa036444682
|
||||
uri: https://huggingface.co/openbmb/MiniCPM-OCR-1.0-gguf/resolve/9ba8c0db38ec333bdd83d44a88d25c6f1d748ba4/mmproj-model-f16-4x.gguf
|
||||
overrides:
|
||||
backend: llama-cpp
|
||||
context_size: 8192
|
||||
known_usecases:
|
||||
- chat
|
||||
- vision
|
||||
mmproj: llama-cpp/models/minicpm-ocr-1.0/mmproj-model-f16-4x.gguf
|
||||
options:
|
||||
- use_jinja:true
|
||||
- enable_reasoning:0
|
||||
- parallel:1
|
||||
parameters:
|
||||
model: llama-cpp/models/minicpm-ocr-1.0/MiniCPM-OCR-1_0-Q8_0.gguf
|
||||
temperature: 0.01
|
||||
top_k: 1
|
||||
top_p: 0.001
|
||||
template:
|
||||
use_tokenizer_template: true
|
||||
variants:
|
||||
- model: minicpm-ocr-1.0-q4km
|
||||
- model: minicpm-ocr-1.0-dspark
|
||||
- url: github:mudler/LocalAI/gallery/virtual.yaml@master
|
||||
name: minicpm-ocr-1.0-q4km
|
||||
description: |
|
||||
MiniCPM-OCR 1.0 in the smaller Q4_K_M GGUF build, with the required F16
|
||||
4x vision projector and an 8,192-token default context. Tables stay
|
||||
lossless at this size; dense text may degrade slightly. Prompt the model
|
||||
with "Text Recognition:" for text, "Table Recognition:" for tables, or
|
||||
"Formula Recognition:" for formulas. Table output is the model's OTSL
|
||||
markup, which clients can convert to HTML.
|
||||
license: apache-2.0
|
||||
urls:
|
||||
- https://huggingface.co/openbmb/MiniCPM-OCR-1.0
|
||||
- https://huggingface.co/openbmb/MiniCPM-OCR-1.0-gguf
|
||||
icon: https://avatars.githubusercontent.com/u/89920203
|
||||
tags:
|
||||
- llm
|
||||
- gguf
|
||||
- gpu
|
||||
- cpu
|
||||
- image-to-text
|
||||
- ocr
|
||||
- document-parsing
|
||||
- multimodal
|
||||
- vision
|
||||
- minicpm
|
||||
files:
|
||||
- filename: llama-cpp/models/minicpm-ocr-1.0/MiniCPM-OCR-1_0-Q4_K_M.gguf
|
||||
sha256: c80a703400633f298cc2fddf6d493a541be36918c0801c0b1bf3f14c5703ce49
|
||||
uri: https://huggingface.co/openbmb/MiniCPM-OCR-1.0-gguf/resolve/9ba8c0db38ec333bdd83d44a88d25c6f1d748ba4/MiniCPM-OCR-1_0-Q4_K_M.gguf
|
||||
- filename: llama-cpp/models/minicpm-ocr-1.0/mmproj-model-f16-4x.gguf
|
||||
sha256: ceed82261b3b1380c3165f8b37c10117a50093be1362cbfeeb692fa036444682
|
||||
uri: https://huggingface.co/openbmb/MiniCPM-OCR-1.0-gguf/resolve/9ba8c0db38ec333bdd83d44a88d25c6f1d748ba4/mmproj-model-f16-4x.gguf
|
||||
overrides:
|
||||
backend: llama-cpp
|
||||
context_size: 8192
|
||||
known_usecases:
|
||||
- chat
|
||||
- vision
|
||||
mmproj: llama-cpp/models/minicpm-ocr-1.0/mmproj-model-f16-4x.gguf
|
||||
options:
|
||||
- use_jinja:true
|
||||
- enable_reasoning:0
|
||||
- parallel:1
|
||||
parameters:
|
||||
model: llama-cpp/models/minicpm-ocr-1.0/MiniCPM-OCR-1_0-Q4_K_M.gguf
|
||||
temperature: 0.01
|
||||
top_k: 1
|
||||
top_p: 0.001
|
||||
template:
|
||||
use_tokenizer_template: true
|
||||
- url: github:mudler/LocalAI/gallery/virtual.yaml@master
|
||||
name: minicpm-ocr-1.0-dspark
|
||||
description: |
|
||||
MiniCPM-OCR 1.0 with OpenBMB's DSpark speculative drafter. DSpark
|
||||
proposes blocks of tokens that the target verifies, which speeds up
|
||||
generation without changing the target model's output. This entry pairs
|
||||
the Q8_0 target with the matching Q8_0 draft and the required F16 4x
|
||||
vision projector, with an 8,192-token default context.
|
||||
license: apache-2.0
|
||||
urls:
|
||||
- https://huggingface.co/openbmb/MiniCPM-OCR-1.0
|
||||
- https://huggingface.co/openbmb/MiniCPM-OCR-1.0-gguf
|
||||
- https://huggingface.co/openbmb/MiniCPM-OCR-1.0-DSpark
|
||||
- https://huggingface.co/openbmb/MiniCPM-OCR-1.0-DSpark-gguf
|
||||
icon: https://avatars.githubusercontent.com/u/89920203
|
||||
tags:
|
||||
- llm
|
||||
- gguf
|
||||
- gpu
|
||||
- cpu
|
||||
- image-to-text
|
||||
- ocr
|
||||
- document-parsing
|
||||
- multimodal
|
||||
- vision
|
||||
- minicpm
|
||||
- dspark
|
||||
files:
|
||||
- filename: llama-cpp/models/minicpm-ocr-1.0/MiniCPM-OCR-1_0-Q8_0.gguf
|
||||
sha256: 791486a476bdacb48acb3f31aa3a18dbb5be87e723a4074bcd13a37d61c852e2
|
||||
uri: https://huggingface.co/openbmb/MiniCPM-OCR-1.0-gguf/resolve/9ba8c0db38ec333bdd83d44a88d25c6f1d748ba4/MiniCPM-OCR-1_0-Q8_0.gguf
|
||||
- filename: llama-cpp/models/minicpm-ocr-1.0/dspark-MiniCPM-OCR-1_0-Q8_0.gguf
|
||||
sha256: fd8f64bb62cf5a5c6b4c0999bfa58f64453690c9f75f1124ab613dfb44a822da
|
||||
uri: https://huggingface.co/openbmb/MiniCPM-OCR-1.0-DSpark-gguf/resolve/265c4dc398c8d5c0ac56a4dd155a5a97b2037226/dspark-MiniCPM-OCR-1_0-Q8_0.gguf
|
||||
- filename: llama-cpp/models/minicpm-ocr-1.0/mmproj-model-f16-4x.gguf
|
||||
sha256: ceed82261b3b1380c3165f8b37c10117a50093be1362cbfeeb692fa036444682
|
||||
uri: https://huggingface.co/openbmb/MiniCPM-OCR-1.0-gguf/resolve/9ba8c0db38ec333bdd83d44a88d25c6f1d748ba4/mmproj-model-f16-4x.gguf
|
||||
overrides:
|
||||
backend: llama-cpp
|
||||
context_size: 8192
|
||||
draft_model: llama-cpp/models/minicpm-ocr-1.0/dspark-MiniCPM-OCR-1_0-Q8_0.gguf
|
||||
known_usecases:
|
||||
- chat
|
||||
- vision
|
||||
mmproj: llama-cpp/models/minicpm-ocr-1.0/mmproj-model-f16-4x.gguf
|
||||
options:
|
||||
- use_jinja:true
|
||||
- enable_reasoning:0
|
||||
- parallel:1
|
||||
- spec_type:draft-dspark
|
||||
- spec_n_max:15
|
||||
parameters:
|
||||
model: llama-cpp/models/minicpm-ocr-1.0/MiniCPM-OCR-1_0-Q8_0.gguf
|
||||
temperature: 0.01
|
||||
top_k: 1
|
||||
top_p: 0.001
|
||||
template:
|
||||
use_tokenizer_template: true
|
||||
- url: github:mudler/LocalAI/gallery/jamba.yaml@master
|
||||
name: ai21labs_ai21-jamba-reasoning-3b
|
||||
description: |
|
||||
|
||||
@@ -36648,6 +36648,192 @@
|
||||
model: OvisOCR2-Q8_0.gguf
|
||||
template:
|
||||
use_tokenizer_template: true
|
||||
- kind: model
|
||||
id: minicpm-ocr-1.0-q8
|
||||
name: minicpm-ocr-1.0-q8
|
||||
version: 0.1.0
|
||||
description: |
|
||||
MiniCPM-OCR 1.0 is OpenBMB's 1.3B document-parsing model for Chinese and
|
||||
English text, tables, formulas, charts, and seals. This entry uses the
|
||||
official Q8_0 GGUF weights and the required F16 4x vision projector
|
||||
through llama.cpp, with an 8,192-token default context. Prompt the model
|
||||
with "Text Recognition:" for text, "Table Recognition:" for tables, or
|
||||
"Formula Recognition:" for formulas. Table output is the model's OTSL
|
||||
markup, which clients can convert to HTML.
|
||||
license: apache-2.0
|
||||
urls:
|
||||
- https://huggingface.co/openbmb/MiniCPM-OCR-1.0
|
||||
- https://huggingface.co/openbmb/MiniCPM-OCR-1.0-gguf
|
||||
tags:
|
||||
- llm
|
||||
- gguf
|
||||
- gpu
|
||||
- cpu
|
||||
- image-to-text
|
||||
- ocr
|
||||
- document-parsing
|
||||
- multimodal
|
||||
- vision
|
||||
- minicpm
|
||||
icon: https://avatars.githubusercontent.com/u/89920203
|
||||
channel: gallery
|
||||
policy:
|
||||
requireSignature: false
|
||||
schemaVersion: 1
|
||||
model:
|
||||
url: github:mudler/LocalAI/gallery/virtual.yaml@master
|
||||
files:
|
||||
- filename: llama-cpp/models/minicpm-ocr-1.0/MiniCPM-OCR-1_0-Q8_0.gguf
|
||||
sha256: 791486a476bdacb48acb3f31aa3a18dbb5be87e723a4074bcd13a37d61c852e2
|
||||
uri: https://huggingface.co/openbmb/MiniCPM-OCR-1.0-gguf/resolve/9ba8c0db38ec333bdd83d44a88d25c6f1d748ba4/MiniCPM-OCR-1_0-Q8_0.gguf
|
||||
- filename: llama-cpp/models/minicpm-ocr-1.0/mmproj-model-f16-4x.gguf
|
||||
sha256: ceed82261b3b1380c3165f8b37c10117a50093be1362cbfeeb692fa036444682
|
||||
uri: https://huggingface.co/openbmb/MiniCPM-OCR-1.0-gguf/resolve/9ba8c0db38ec333bdd83d44a88d25c6f1d748ba4/mmproj-model-f16-4x.gguf
|
||||
overrides:
|
||||
backend: llama-cpp
|
||||
context_size: 8192
|
||||
known_usecases:
|
||||
- chat
|
||||
- vision
|
||||
mmproj: llama-cpp/models/minicpm-ocr-1.0/mmproj-model-f16-4x.gguf
|
||||
options:
|
||||
- use_jinja:true
|
||||
- enable_reasoning:0
|
||||
- parallel:1
|
||||
parameters:
|
||||
model: llama-cpp/models/minicpm-ocr-1.0/MiniCPM-OCR-1_0-Q8_0.gguf
|
||||
temperature: 0.01
|
||||
top_p: 0.001
|
||||
top_k: 1
|
||||
template:
|
||||
use_tokenizer_template: true
|
||||
variants:
|
||||
- model: minicpm-ocr-1.0-q4km
|
||||
- model: minicpm-ocr-1.0-dspark
|
||||
- kind: model
|
||||
id: minicpm-ocr-1.0-q4km
|
||||
name: minicpm-ocr-1.0-q4km
|
||||
version: 0.1.0
|
||||
description: |
|
||||
MiniCPM-OCR 1.0 in the smaller Q4_K_M GGUF build, with the required F16
|
||||
4x vision projector and an 8,192-token default context. Tables stay
|
||||
lossless at this size; dense text may degrade slightly. Prompt the model
|
||||
with "Text Recognition:" for text, "Table Recognition:" for tables, or
|
||||
"Formula Recognition:" for formulas. Table output is the model's OTSL
|
||||
markup, which clients can convert to HTML.
|
||||
license: apache-2.0
|
||||
urls:
|
||||
- https://huggingface.co/openbmb/MiniCPM-OCR-1.0
|
||||
- https://huggingface.co/openbmb/MiniCPM-OCR-1.0-gguf
|
||||
tags:
|
||||
- llm
|
||||
- gguf
|
||||
- gpu
|
||||
- cpu
|
||||
- image-to-text
|
||||
- ocr
|
||||
- document-parsing
|
||||
- multimodal
|
||||
- vision
|
||||
- minicpm
|
||||
icon: https://avatars.githubusercontent.com/u/89920203
|
||||
channel: gallery
|
||||
policy:
|
||||
requireSignature: false
|
||||
schemaVersion: 1
|
||||
model:
|
||||
url: github:mudler/LocalAI/gallery/virtual.yaml@master
|
||||
files:
|
||||
- filename: llama-cpp/models/minicpm-ocr-1.0/MiniCPM-OCR-1_0-Q4_K_M.gguf
|
||||
sha256: c80a703400633f298cc2fddf6d493a541be36918c0801c0b1bf3f14c5703ce49
|
||||
uri: https://huggingface.co/openbmb/MiniCPM-OCR-1.0-gguf/resolve/9ba8c0db38ec333bdd83d44a88d25c6f1d748ba4/MiniCPM-OCR-1_0-Q4_K_M.gguf
|
||||
- filename: llama-cpp/models/minicpm-ocr-1.0/mmproj-model-f16-4x.gguf
|
||||
sha256: ceed82261b3b1380c3165f8b37c10117a50093be1362cbfeeb692fa036444682
|
||||
uri: https://huggingface.co/openbmb/MiniCPM-OCR-1.0-gguf/resolve/9ba8c0db38ec333bdd83d44a88d25c6f1d748ba4/mmproj-model-f16-4x.gguf
|
||||
overrides:
|
||||
backend: llama-cpp
|
||||
context_size: 8192
|
||||
known_usecases:
|
||||
- chat
|
||||
- vision
|
||||
mmproj: llama-cpp/models/minicpm-ocr-1.0/mmproj-model-f16-4x.gguf
|
||||
options:
|
||||
- use_jinja:true
|
||||
- enable_reasoning:0
|
||||
- parallel:1
|
||||
parameters:
|
||||
model: llama-cpp/models/minicpm-ocr-1.0/MiniCPM-OCR-1_0-Q4_K_M.gguf
|
||||
temperature: 0.01
|
||||
top_p: 0.001
|
||||
top_k: 1
|
||||
template:
|
||||
use_tokenizer_template: true
|
||||
- kind: model
|
||||
id: minicpm-ocr-1.0-dspark
|
||||
name: minicpm-ocr-1.0-dspark
|
||||
version: 0.1.0
|
||||
description: |
|
||||
MiniCPM-OCR 1.0 with OpenBMB's DSpark speculative drafter. DSpark
|
||||
proposes blocks of tokens that the target verifies, which speeds up
|
||||
generation without changing the target model's output. This entry pairs
|
||||
the Q8_0 target with the matching Q8_0 draft and the required F16 4x
|
||||
vision projector, with an 8,192-token default context.
|
||||
license: apache-2.0
|
||||
urls:
|
||||
- https://huggingface.co/openbmb/MiniCPM-OCR-1.0
|
||||
- https://huggingface.co/openbmb/MiniCPM-OCR-1.0-gguf
|
||||
- https://huggingface.co/openbmb/MiniCPM-OCR-1.0-DSpark
|
||||
- https://huggingface.co/openbmb/MiniCPM-OCR-1.0-DSpark-gguf
|
||||
tags:
|
||||
- llm
|
||||
- gguf
|
||||
- gpu
|
||||
- cpu
|
||||
- image-to-text
|
||||
- ocr
|
||||
- document-parsing
|
||||
- multimodal
|
||||
- vision
|
||||
- minicpm
|
||||
- dspark
|
||||
icon: https://avatars.githubusercontent.com/u/89920203
|
||||
channel: gallery
|
||||
policy:
|
||||
requireSignature: false
|
||||
schemaVersion: 1
|
||||
model:
|
||||
url: github:mudler/LocalAI/gallery/virtual.yaml@master
|
||||
files:
|
||||
- filename: llama-cpp/models/minicpm-ocr-1.0/MiniCPM-OCR-1_0-Q8_0.gguf
|
||||
sha256: 791486a476bdacb48acb3f31aa3a18dbb5be87e723a4074bcd13a37d61c852e2
|
||||
uri: https://huggingface.co/openbmb/MiniCPM-OCR-1.0-gguf/resolve/9ba8c0db38ec333bdd83d44a88d25c6f1d748ba4/MiniCPM-OCR-1_0-Q8_0.gguf
|
||||
- filename: llama-cpp/models/minicpm-ocr-1.0/dspark-MiniCPM-OCR-1_0-Q8_0.gguf
|
||||
sha256: fd8f64bb62cf5a5c6b4c0999bfa58f64453690c9f75f1124ab613dfb44a822da
|
||||
uri: https://huggingface.co/openbmb/MiniCPM-OCR-1.0-DSpark-gguf/resolve/265c4dc398c8d5c0ac56a4dd155a5a97b2037226/dspark-MiniCPM-OCR-1_0-Q8_0.gguf
|
||||
- filename: llama-cpp/models/minicpm-ocr-1.0/mmproj-model-f16-4x.gguf
|
||||
sha256: ceed82261b3b1380c3165f8b37c10117a50093be1362cbfeeb692fa036444682
|
||||
uri: https://huggingface.co/openbmb/MiniCPM-OCR-1.0-gguf/resolve/9ba8c0db38ec333bdd83d44a88d25c6f1d748ba4/mmproj-model-f16-4x.gguf
|
||||
overrides:
|
||||
backend: llama-cpp
|
||||
context_size: 8192
|
||||
draft_model: llama-cpp/models/minicpm-ocr-1.0/dspark-MiniCPM-OCR-1_0-Q8_0.gguf
|
||||
known_usecases:
|
||||
- chat
|
||||
- vision
|
||||
mmproj: llama-cpp/models/minicpm-ocr-1.0/mmproj-model-f16-4x.gguf
|
||||
options:
|
||||
- use_jinja:true
|
||||
- enable_reasoning:0
|
||||
- parallel:1
|
||||
- spec_type:draft-dspark
|
||||
- spec_n_max:15
|
||||
parameters:
|
||||
model: llama-cpp/models/minicpm-ocr-1.0/MiniCPM-OCR-1_0-Q8_0.gguf
|
||||
temperature: 0.01
|
||||
top_p: 0.001
|
||||
top_k: 1
|
||||
template:
|
||||
use_tokenizer_template: true
|
||||
- kind: model
|
||||
id: ai21labs_ai21-jamba-reasoning-3b
|
||||
name: ai21labs_ai21-jamba-reasoning-3b
|
||||
|
||||
Reference in new issue
Block a user