注意：MistralRS需要安装名为cargo的Rust包管理器。请访问https://rustup.rs/查看安装详情。

In [ ]

已复制！

%pip install llama-index-core
%pip install llama-index-readers-file
%pip install llama-index-llms-mistral-rs
%pip install llama-index-llms-huggingface

from llama_index.core import VectorStoreIndex, SimpleDirectoryReader, Settings from llama_index.core.embeddings import resolve_embed_model from llama_index.llms.mistral_rs import MistralRS from mistralrs import Which, Architecture documents = SimpleDirectoryReader("data").load_data() # bge embedding model Settings.embed_model = resolve_embed_model("local:BAAI/bge-small-en-v1.5")

已复制！

from llama_index.core import VectorStoreIndex, SimpleDirectoryReader, Settings
from llama_index.core.embeddings import resolve_embed_model
from llama_index.llms.mistral_rs import MistralRS
from mistralrs import Which, Architecture

documents = SimpleDirectoryReader("data").load_data()

# bge embedding model
Settings.embed_model = resolve_embed_model("local:BAAI/bge-small-en-v1.5")

MistralRS使用huggingface hub中的模型ID。

# Full Model Settings.llm = MistralRS( which=Which.Plain( model_id="mistralai/Mistral-7B-Instruct-v0.1", arch=Architecture.Mistral, tokenizer_json=None, repeat_last_n=64, ), max_new_tokens=4096, context_window=1024 * 5, ) # GGUF Model, Quantized Settings.llm = MistralRS( which=Which.GGUF( tok_model_id="mistralai/Mistral-7B-Instruct-v0.1", quantized_model_id="TheBloke/Mistral-7B-Instruct-v0.1-GGUF", quantized_filename="mistral-7b-instruct-v0.1.Q4_K_M.gguf", tokenizer_json=None, repeat_last_n=64, ), max_new_tokens=4096, context_window=1024 * 5, )

已复制！

# Full Model
Settings.llm = MistralRS(
    which=Which.Plain(
        model_id="mistralai/Mistral-7B-Instruct-v0.1",
        arch=Architecture.Mistral,
        tokenizer_json=None,
        repeat_last_n=64,
    ),
    max_new_tokens=4096,
    context_window=1024 * 5,
)

# GGUF Model, Quantized
Settings.llm = MistralRS(
    which=Which.GGUF(
        tok_model_id="mistralai/Mistral-7B-Instruct-v0.1",
        quantized_model_id="TheBloke/Mistral-7B-Instruct-v0.1-GGUF",
        quantized_filename="mistral-7b-instruct-v0.1.Q4_K_M.gguf",
        tokenizer_json=None,
        repeat_last_n=64,
    ),
    max_new_tokens=4096,
    context_window=1024 * 5,
)

index = VectorStoreIndex.from_documents( documents, )

已复制！

index = VectorStoreIndex.from_documents(
    documents,
)

query_engine = index.as_query_engine() response = query_engine.query("How do I pronounce graphene?") print(response)

已复制！

query_engine = index.as_query_engine()
response = query_engine.query("How do I pronounce graphene?")
print(response)

回到顶部