add rerank model (#969)

### What problem does this PR solve?

feat: add rerank models to the project #724 #162

### Type of change

- [x] New Feature (non-breaking change which adds functionality)
This commit is contained in:
KevinHuSh
2024-05-29 16:50:02 +08:00
committed by GitHub
parent e1f0644deb
commit 614defec21
17 changed files with 437 additions and 64 deletions

View File

@ -142,7 +142,17 @@ factory_infos = [{
"logo": "",
"tags": "LLM,TEXT EMBEDDING",
"status": "1",
},
},{
"name": "Jina",
"logo": "",
"tags": "TEXT EMBEDDING, TEXT RE-RANK",
"status": "1",
},{
"name": "BAAI",
"logo": "",
"tags": "TEXT EMBEDDING, TEXT RE-RANK",
"status": "1",
}
# {
# "name": "文心一言",
# "logo": "",
@ -367,6 +377,13 @@ def init_llm_factory():
"max_tokens": 512,
"model_type": LLMType.EMBEDDING.value
},
{
"fid": factory_infos[7]["name"],
"llm_name": "maidalun1020/bce-reranker-base_v1",
"tags": "RE-RANK, 8K",
"max_tokens": 8196,
"model_type": LLMType.RERANK.value
},
# ------------------------ DeepSeek -----------------------
{
"fid": factory_infos[8]["name"],
@ -440,6 +457,85 @@ def init_llm_factory():
"max_tokens": 512,
"model_type": LLMType.EMBEDDING.value
},
# ------------------------ Jina -----------------------
{
"fid": factory_infos[11]["name"],
"llm_name": "jina-reranker-v1-base-en",
"tags": "RE-RANK,8k",
"max_tokens": 8196,
"model_type": LLMType.RERANK.value
},
{
"fid": factory_infos[11]["name"],
"llm_name": "jina-reranker-v1-turbo-en",
"tags": "RE-RANK,8k",
"max_tokens": 8196,
"model_type": LLMType.RERANK.value
},
{
"fid": factory_infos[11]["name"],
"llm_name": "jina-reranker-v1-tiny-en",
"tags": "RE-RANK,8k",
"max_tokens": 8196,
"model_type": LLMType.RERANK.value
},
{
"fid": factory_infos[11]["name"],
"llm_name": "jina-colbert-v1-en",
"tags": "RE-RANK,8k",
"max_tokens": 8196,
"model_type": LLMType.RERANK.value
},
{
"fid": factory_infos[11]["name"],
"llm_name": "jina-embeddings-v2-base-en",
"tags": "TEXT EMBEDDING",
"max_tokens": 8196,
"model_type": LLMType.EMBEDDING.value
},
{
"fid": factory_infos[11]["name"],
"llm_name": "jina-embeddings-v2-base-de",
"tags": "TEXT EMBEDDING",
"max_tokens": 8196,
"model_type": LLMType.EMBEDDING.value
},
{
"fid": factory_infos[11]["name"],
"llm_name": "jina-embeddings-v2-base-es",
"tags": "TEXT EMBEDDING",
"max_tokens": 8196,
"model_type": LLMType.EMBEDDING.value
},
{
"fid": factory_infos[11]["name"],
"llm_name": "jina-embeddings-v2-base-code",
"tags": "TEXT EMBEDDING",
"max_tokens": 8196,
"model_type": LLMType.EMBEDDING.value
},
{
"fid": factory_infos[11]["name"],
"llm_name": "jina-embeddings-v2-base-zh",
"tags": "TEXT EMBEDDING",
"max_tokens": 8196,
"model_type": LLMType.EMBEDDING.value
},
# ------------------------ BAAI -----------------------
{
"fid": factory_infos[12]["name"],
"llm_name": "BAAI/bge-large-zh-v1.5",
"tags": "TEXT EMBEDDING,",
"max_tokens": 1024,
"model_type": LLMType.EMBEDDING.value
},
{
"fid": factory_infos[12]["name"],
"llm_name": "BAAI/bge-reranker-v2-m3",
"tags": "LLM,CHAT,",
"max_tokens": 16385,
"model_type": LLMType.RERANK.value
},
]
for info in factory_infos:
try: