jealk/TTC-unsupervised-1

The Tech Collective - Unsupervised Embedding model (Danish)

Model Description

Unsupervised model for sentence embeddings.

• Developed by: Jesper Alkestrup, The Tech Collective
• Model type: Embedding model
• Language(s) (NLP): Danish
• Finetuned from model : AI-Sweden-Models/Llama-3-8B-instruct
• Finetuning procedure: LLM2Vec

Trained by using the approach outlined in the paper LLM2Vec: Large Language Models Are Secretly Powerful Text Encoders.

LoRa Finetuning 1000 steps of MNTP on cleaned Danish Wikipedia https://huggingface.co/datasets/jealk/wiki40b-da-clean LoRa Finetuning 1000 steps of SimCSE on sentences from Scandivian Wikipedia (da, nn, nb, sv, fo, is): https://huggingface.co/datasets/jealk/scandi-wiki-combined (Sentence subset)

Credits for code-repo used to finetune this model https://github.com/McGill-NLP/llm2vec

Requires the llm2vec package to encode sentences. Credits to https://huggingface.co/McGill-NLP/LLM2Vec-Meta-Llama-3-8B-Instruct-mntp-supervised for the below instructions:

Installation

pip install llm2vec

Usage

from llm2vec import LLM2Vec

import torch
from transformers import AutoTokenizer, AutoModel, AutoConfig
from peft import PeftModel

# Loading base Mistral model, along with custom code that enables bidirectional connections in decoder-only LLMs. MNTP LoRA weights are merged into the base model.
tokenizer = AutoTokenizer.from_pretrained(
    "jealk/llm2vec-da-mntp"
)
config = AutoConfig.from_pretrained(
    "jealk/llm2vec-da-mntp", trust_remote_code=True
)
model = AutoModel.from_pretrained(
    "jealk/llm2vec-da-mntp",
    trust_remote_code=True,
    config=config,
    torch_dtype=torch.bfloat16,
    device_map="cuda" if torch.cuda.is_available() else "cpu",
)
model = PeftModel.from_pretrained(
    model,
    "jealk/llm2vec-da-mntp",
)
model = model.merge_and_unload()  # This can take several minutes on cpu

# Loading supervised model. This loads the trained LoRA weights on top of MNTP model. Hence the final weights are -- Base model + MNTP (LoRA) + supervised (LoRA).
model = PeftModel.from_pretrained(
    model, "jealk/TTC-unsupervised-1"
)

# Wrapper for encoding and pooling operations
l2v = LLM2Vec(model, tokenizer, pooling_mode="mean", max_length=512)

# Encoding queries using instructions
instruction = (
    "Given a web search query, retrieve relevant passages that answer the query:"
)
queries = [
    [instruction, "how much protein should a female eat"],
    [instruction, "summit define"],
]
q_reps = l2v.encode(queries)

# Encoding documents. Instruction are not required for documents
documents = [
    "As a general guideline, the CDC's average requirement of protein for women ages 19 to 70 is 46 grams per day. But, as you can see from this chart, you'll need to increase that if you're expecting or training for a marathon. Check out the chart below to see how much protein you should be eating each day.",
    "Definition of summit for English Language Learners. : 1  the highest point of a mountain : the top of a mountain. : 2  the highest level. : 3  a meeting or series of meetings between the leaders of two or more governments.",
]
d_reps = l2v.encode(documents)

# Compute cosine similarity
q_reps_norm = torch.nn.functional.normalize(q_reps, p=2, dim=1)
d_reps_norm = torch.nn.functional.normalize(d_reps, p=2, dim=1)
cos_sim = torch.mm(q_reps_norm, d_reps_norm.transpose(0, 1))

print(cos_sim)
"""
tensor([[0.6470, 0.1619],
        [0.0786, 0.5844]])
"""