You can deploy this model and start earning money today!

Rakesh44/
odyssey-fin-sentiment

Name: odyssey-fin-sentiment
Availability: InStock

peft

safetensors

lora

sequence-classification

Views

No views yet

https://huggingface.co/Rakesh44/odyssey-fin-sentiment (LoRA Adapter — Sequence Classification)

A LoRA adapter fine-tuned on top of Rakesh44/odyssey for [TASK — e.g. financial sentiment analysis / market sentiment classification]. This repository contains the adapter weights only; the base model is loaded separately and the adapter is applied on top.

Model Details

Developed by: Rakesh44
**Base model: https://huggingface.co/Rakesh44/odyssey
Task: Sequence classification (SEQ_CLS)
Number of labels: positive / neutral / negative
Adapter type: LoRA (PEFT)
Language: English
License: Apache 2.0

LoRA Configuration

Setting	Value
PEFT type	LoRA
Rank (r)	16
Alpha	32
Dropout	0.05
Bias	none
Target modules	q_proj, k_proj, v_proj, o_proj
Modules saved (full)	score / classifier head
RSLoRA / DoRA / QALoRA	disabled
PEFT version	0.19.1

The adapter targets the attention projections only; the classification head (score/classifier) is trained in full and saved with the adapter.

Uses

Intended use

Classifying the sentiment of financial or market text — headlines, reports, commentary — into fixed categories.

Out of scope

Not intended for text generation, question answering, or high-stakes automated decisions. Predictions on financial or market text should be independently validated before any real-world use.

How to Use

python

from transformers import AutoTokenizer, AutoModelForSequenceClassification
from peft import PeftModel
import torch

base_id = "Rakesh44/odyssey"
adapter_id = "Rakesh44/odyssey-fin-sentiment"

tokenizer = AutoTokenizer.from_pretrained(base_id)
base = AutoModelForSequenceClassification.from_pretrained(
    base_id, num_labels=[NUM_LABELS]
)
model = PeftModel.from_pretrained(base, adapter_id)
model.eval()

text = "The company beat earnings expectations this quarter."
inputs = tokenizer(text, return_tensors="pt", truncation=True)
with torch.no_grad():
    logits = model(**inputs).logits
pred = logits.argmax(-1).item()
print(model.config.id2label[pred])

Training

Method: LoRA fine-tuning (attention projections) with a fully-trained classification head
Hardware: 1× GPU