metadata

license: apache-2.0
tags:
  - fine-tuned
  - qwen2.5c
  - instruct
  - causal-lm

Qwen2.5-7b-instruct-sft

This is a fine-tuned version of the Qwen2.5-7B-Instruct model for instruction-following tasks. It was fine-tuned using the SFTTrainer from the trl library on the OpenAssistant Guanaco dataset.

Model Details

Base Model

Model: Qwen2.5-7B-Instruct
Architecture: Transformer-based causal language model
License: Apache 2.0

Fine-Tuning Details

Dataset: OpenAssistant Guanaco
Training Epochs: 1
Batch Size: 2
Gradient Accumulation Steps: 16
Learning Rate: 1e-5
Optimizer: Paged AdamW 8-bit
Mixed Precision: fp16 (if bf16 is not supported) or bf16
Max Sequence Length: 512 tokens

Training Hardware

GPU: NVIDIA A100 (or your specific GPU)
Training Time: X hours (optional)

Usage

You can use this model with the Hugging Face transformers library:

from transformers import AutoModelForCausalLM, AutoTokenizer

# Load the model and tokenizer
model = AutoModelForCausalLM.from_pretrained("Ameer-linnk/qwen2.5-7b-instruct-sft")
tokenizer = AutoTokenizer.from_pretrained("Ameer-linnk/qwen2.5-7b-instruct-sft")

# Prepare input
input_text = "What is the capital of France?"
inputs = tokenizer(input_text, return_tensors="pt")

# Generate output
outputs = model.generate(**inputs)
response = tokenizer.decode(outputs[0], skip_special_tokens=True)

print(response) 
Example Output
Input:
What is the capital of France?

Output:
The capital of France is Paris.