mirror of
https://github.com/nomic-ai/gpt4all.git
synced 2025-04-28 03:41:58 +00:00
models: add DeepSeek-R1 distillations to official models list (#3437)
Signed-off-by: Jared Van Bortel <jared@nomic.ai>
This commit is contained in:
Jared Van Bortel
GitHub
parent
007a7af1c8
commit
34037f3101
@ -9,6 +9,7 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/).
|
||||
### Added
|
||||
- Support DeepSeek-R1 Qwen models ([#3431](https://github.com/nomic-ai/gpt4all/pull/3431))
|
||||
- Support for think tags in the GUI ([#3440](https://github.com/nomic-ai/gpt4all/pull/3440))
|
||||
- Support specifying SHA256 hash in models3.json instead of MD5 ([#3437](https://github.com/nomic-ai/gpt4all/pull/3437))
|
||||
|
||||
### Changed
|
||||
- Use minja instead of Jinja2Cpp for significantly improved template compatibility ([#3433](https://github.com/nomic-ai/gpt4all/pull/3433))
|
||||
|
||||
@ -32,6 +32,66 @@
|
||||
"systemPrompt": "",
|
||||
"chatTemplate": "{%- set loop_messages = messages %}\n{%- for message in loop_messages %}\n {%- set content = '<|start_header_id|>' + message['role'] + '<|end_header_id|>\\n\\n'+ message['content'] | trim + '<|eot_id|>' %}\n {{- content }}\n{%- endfor %}\n{%- if add_generation_prompt %}\n {{- '<|start_header_id|>assistant<|end_header_id|>\\n\\n' }}\n{%- endif %}"
|
||||
},
|
||||
{
|
||||
"order": "aa1",
|
||||
"sha256sum": "5cd4ee65211770f1d99b4f6f4951780b9ef40e29314bd6542bb5bd0ad0bc29d1",
|
||||
"name": "DeepSeek-R1-Distill-Qwen-7B",
|
||||
"filename": "DeepSeek-R1-Distill-Qwen-7B-Q4_0.gguf",
|
||||
"filesize": "4444121056",
|
||||
"requires": "3.8.0",
|
||||
"ramrequired": "8",
|
||||
"parameters": "7 billion",
|
||||
"quant": "q4_0",
|
||||
"type": "deepseek",
|
||||
"description": "<p>The official Qwen2.5-Math-7B distillation of DeepSeek-R1.</p><ul><li>License: <a href=\"https://opensource.org/license/mit\">MIT</a></li><li>No restrictions on commercial use</li><li>#reasoning</li></ul>",
|
||||
"url": "https://huggingface.co/bartowski/DeepSeek-R1-Distill-Qwen-7B-GGUF/resolve/main/DeepSeek-R1-Distill-Qwen-7B-Q4_0.gguf",
|
||||
"chatTemplate": "{%- if not add_generation_prompt is defined %}\n {%- set add_generation_prompt = false %}\n{%- endif %}\n{%- if messages[0]['role'] == 'system' %}\n {{- messages[0]['content'] }}\n{%- endif %}\n{%- for message in messages %}\n {%- if message['role'] == 'user' %}\n {{- '<｜User｜>' + message['content'] }}\n {%- endif %}\n {%- if message['role'] == 'assistant' %}\n {%- set content = message['content'] | regex_replace('^[\\\\s\\\\S]*</think>', '') %}\n {{- '<｜Assistant｜>' + content + '<｜end▁of▁sentence｜>' }}\n {%- endif %}\n{%- endfor -%}\n{%- if add_generation_prompt %}\n {{- '<｜Assistant｜>' }}\n{%- endif %}"
|
||||
},
|
||||
{
|
||||
"order": "aa2",
|
||||
"sha256sum": "906b3382f2680f4ce845459b4a122e904002b075238080307586bcffcde49eef",
|
||||
"name": "DeepSeek-R1-Distill-Qwen-14B",
|
||||
"filename": "DeepSeek-R1-Distill-Qwen-14B-Q4_0.gguf",
|
||||
"filesize": "8544267680",
|
||||
"requires": "3.8.0",
|
||||
"ramrequired": "16",
|
||||
"parameters": "14 billion",
|
||||
"quant": "q4_0",
|
||||
"type": "deepseek",
|
||||
"description": "<p>The official Qwen2.5-14B distillation of DeepSeek-R1.</p><ul><li>License: <a href=\"https://opensource.org/license/mit\">MIT</a></li><li>No restrictions on commercial use</li><li>#reasoning</li></ul>",
|
||||
"url": "https://huggingface.co/bartowski/DeepSeek-R1-Distill-Qwen-14B-GGUF/resolve/main/DeepSeek-R1-Distill-Qwen-14B-Q4_0.gguf",
|
||||
"chatTemplate": "{%- if not add_generation_prompt is defined %}\n {%- set add_generation_prompt = false %}\n{%- endif %}\n{%- if messages[0]['role'] == 'system' %}\n {{- messages[0]['content'] }}\n{%- endif %}\n{%- for message in messages %}\n {%- if message['role'] == 'user' %}\n {{- '<｜User｜>' + message['content'] }}\n {%- endif %}\n {%- if message['role'] == 'assistant' %}\n {%- set content = message['content'] | regex_replace('^[\\\\s\\\\S]*</think>', '') %}\n {{- '<｜Assistant｜>' + content + '<｜end▁of▁sentence｜>' }}\n {%- endif %}\n{%- endfor -%}\n{%- if add_generation_prompt %}\n {{- '<｜Assistant｜>' }}\n{%- endif %}"
|
||||
},
|
||||
{
|
||||
"order": "aa3",
|
||||
"sha256sum": "0eb93e436ac8beec18aceb958c120d282cb2cf5451b23185e7be268fe9d375cc",
|
||||
"name": "DeepSeek-R1-Distill-Llama-8B",
|
||||
"filename": "DeepSeek-R1-Distill-Llama-8B-Q4_0.gguf",
|
||||
"filesize": "4675894112",
|
||||
"requires": "3.8.0",
|
||||
"ramrequired": "8",
|
||||
"parameters": "8 billion",
|
||||
"quant": "q4_0",
|
||||
"type": "deepseek",
|
||||
"description": "<p>The official Llama-3.1-8B distillation of DeepSeek-R1.</p><ul><li>License: <a href=\"https://opensource.org/license/mit\">MIT</a></li><li>No restrictions on commercial use</li><li>#reasoning</li></ul>",
|
||||
"url": "https://huggingface.co/bartowski/DeepSeek-R1-Distill-Llama-8B-GGUF/resolve/main/DeepSeek-R1-Distill-Llama-8B-Q4_0.gguf",
|
||||
"chatTemplate": "{%- if not add_generation_prompt is defined %}\n {%- set add_generation_prompt = false %}\n{%- endif %}\n{%- if messages[0]['role'] == 'system' %}\n {{- messages[0]['content'] }}\n{%- endif %}\n{%- for message in messages %}\n {%- if message['role'] == 'user' %}\n {{- '<｜User｜>' + message['content'] }}\n {%- endif %}\n {%- if message['role'] == 'assistant' %}\n {%- set content = message['content'] | regex_replace('^[\\\\s\\\\S]*</think>', '') %}\n {{- '<｜Assistant｜>' + content + '<｜end▁of▁sentence｜>' }}\n {%- endif %}\n{%- endfor -%}\n{%- if add_generation_prompt %}\n {{- '<｜Assistant｜>' }}\n{%- endif %}"
|
||||
},
|
||||
{
|
||||
"order": "aa4",
|
||||
"sha256sum": "b3af887d0a015b39fab2395e4faf682c1a81a6a3fd09a43f0d4292f7d94bf4d0",
|
||||
"name": "DeepSeek-R1-Distill-Qwen-1.5B",
|
||||
"filename": "DeepSeek-R1-Distill-Qwen-1.5B-Q4_0.gguf",
|
||||
"filesize": "1068807776",
|
||||
"requires": "3.8.0",
|
||||
"ramrequired": "3",
|
||||
"parameters": "1.5 billion",
|
||||
"quant": "q4_0",
|
||||
"type": "deepseek",
|
||||
"description": "<p>The official Qwen2.5-Math-1.5B distillation of DeepSeek-R1.</p><ul><li>License: <a href=\"https://opensource.org/license/mit\">MIT</a></li><li>No restrictions on commercial use</li><li>#reasoning</li></ul>",
|
||||
"url": "https://huggingface.co/bartowski/DeepSeek-R1-Distill-Qwen-1.5B-GGUF/resolve/main/DeepSeek-R1-Distill-Qwen-1.5B-Q4_0.gguf",
|
||||
"chatTemplate": "{%- if not add_generation_prompt is defined %}\n {%- set add_generation_prompt = false %}\n{%- endif %}\n{%- if messages[0]['role'] == 'system' %}\n {{- messages[0]['content'] }}\n{%- endif %}\n{%- for message in messages %}\n {%- if message['role'] == 'user' %}\n {{- '<｜User｜>' + message['content'] }}\n {%- endif %}\n {%- if message['role'] == 'assistant' %}\n {%- set content = message['content'] | regex_replace('^[\\\\s\\\\S]*</think>', '') %}\n {{- '<｜Assistant｜>' + content + '<｜end▁of▁sentence｜>' }}\n {%- endif %}\n{%- endfor -%}\n{%- if add_generation_prompt %}\n {{- '<｜Assistant｜>' }}\n{%- endif %}"
|
||||
},
|
||||
{
|
||||
"order": "b",
|
||||
"md5sum": "27b44e8ae1817525164ddf4f8dae8af4",
|
||||
@ -472,7 +532,7 @@
|
||||
"filename": "qwen2-1_5b-instruct-q4_0.gguf",
|
||||
"filesize": "937532800",
|
||||
"requires": "3.0",
|
||||
"ramrequired": "4",
|
||||
"ramrequired": "3",
|
||||
"parameters": "1.5 billion",
|
||||
"quant": "q4_0",
|
||||
"type": "qwen2",
|
||||
|
||||
@ -1621,7 +1621,6 @@ void ModelList::parseModelsJsonFile(const QByteArray &jsonData, bool save)
|
||||
QString requiresVersion = obj["requires"].toString();
|
||||
QString versionRemoved = obj["removedIn"].toString();
|
||||
QString url = obj["url"].toString();
|
||||
QByteArray modelHash = obj["md5sum"].toString().toLatin1();
|
||||
bool isDefault = obj.contains("isDefault") && obj["isDefault"] == u"true"_s;
|
||||
bool disableGUI = obj.contains("disableGUI") && obj["disableGUI"] == u"true"_s;
|
||||
QString description = obj["description"].toString();
|
||||
@ -1632,6 +1631,16 @@ void ModelList::parseModelsJsonFile(const QByteArray &jsonData, bool save)
|
||||
QString type = obj["type"].toString();
|
||||
bool isEmbeddingModel = obj["embeddingModel"].toBool();
|
||||
|
||||
QByteArray modelHash;
|
||||
ModelInfo::HashAlgorithm hashAlgorithm;
|
||||
if (auto it = obj.find("sha256sum"_L1); it != obj.end()) {
|
||||
modelHash = it->toString().toLatin1();
|
||||
hashAlgorithm = ModelInfo::Sha256;
|
||||
} else {
|
||||
modelHash = obj["md5sum"].toString().toLatin1();
|
||||
hashAlgorithm = ModelInfo::Md5;
|
||||
}
|
||||
|
||||
// Some models aren't supported in the GUI at all
|
||||
if (disableGUI)
|
||||
continue;
|
||||
@ -1660,7 +1669,7 @@ void ModelList::parseModelsJsonFile(const QByteArray &jsonData, bool save)
|
||||
{ ModelList::FilenameRole, modelFilename },
|
||||
{ ModelList::FilesizeRole, modelFilesize },
|
||||
{ ModelList::HashRole, modelHash },
|
||||
{ ModelList::HashAlgorithmRole, ModelInfo::Md5 },
|
||||
{ ModelList::HashAlgorithmRole, hashAlgorithm },
|
||||
{ ModelList::DefaultRole, isDefault },
|
||||
{ ModelList::DescriptionRole, description },
|
||||
{ ModelList::RequiresVersionRole, requiresVersion },
|
||||
|
||||
Loading…
Reference in New Issue
Block a user