from peft import PeftModel, PeftConfig
from transformers import AutoModelForCausalLM
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline

config = PeftConfig.from_pretrained("shujatoor/phi3nedtuned-ner-json")
model = AutoModelForCausalLM.from_pretrained(
    "microsoft/Phi-3-mini-4k-instruct", 
    device_map="cuda", 
    torch_dtype="auto", 
    trust_remote_code=True, 
)
model = PeftModel.from_pretrained(model, "shujatoor/phi3nedtuned-ner-json")
model.config.to_json_file('adapter_config.json')


torch.random.manual_seed(0)
tokenizer = AutoTokenizer.from_pretrained("shujatoor/phi3nedtuned-ner-json")


text = "Tehzeeb Bakers STRN3277876134234 Block A. Police Foundation,PwD Islamabad 051-5170713-4.051-5170501 STRN#3277876134234 NTN#7261076-2 Sales Receipt 05/04/202405:56:40PM CashierM J Payment:Cash Rate Qty. Total # Descriptlon 80.512.000 190.00 1.VEGETABLESAMOSA Sub Total 161.02 Total Tax: 28.98 POS Service Fee 1.00 Total 191.00 Cash 200.00 Change Due 9.00 SR#th007-220240405175640730 Goods Once Sold Can Not Be Taken Back or Replaced All Prices Are Inclusive Sales Tax 134084240405175640553"
q_json = "extracted_data': {'store_name': '', 'address': '', 'receipt_number': '', 'drug_license_number': '', 'gst_number': '', 'vat_number': '', 'date': '', 'time': '', 'items': [], 'total_items': '', 'gst_tax': '', 'vat_tax': '', 'gross_total': '', 'discount': '', 'net_total': '', 'contact': ''}"
qs = f'{text}. {q_json}'
print('Question:',qs, '\n')
messages = [
    #{"role": "system", "content": ""},
    {"role": "user", "content": qs},

]

pipe = pipeline(
    "text-generation",
    model=model,
    tokenizer=tokenizer,
)

generation_args = {
    "max_new_tokens": 512,
    "return_full_text": False,
    #"temperature": 0.0,
    "do_sample": False,
}

output = pipe(messages, **generation_args)

print('Answer:', output[0]['generated_text'], '\n')

Model description

More information needed

Intended uses & limitations

Named Entity Recognition (NER)

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 0.0002
train_batch_size: 1
eval_batch_size: 1
seed: 0
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: cosine
lr_scheduler_warmup_ratio: 0.2
num_epochs: 5

Training results

Training Loss	Epoch	Step	Validation Loss
1.1904	0.5618	500	1.0617
0.765	1.1236	1000	0.9442
0.782	1.6854	1500	0.8690
0.5591	2.2472	2000	0.8647
0.5669	2.8090	2500	0.8296
0.4205	3.3708	3000	0.8820
0.3812	3.9326	3500	0.8859
0.3323	4.4944	4000	0.9360

Framework versions

PEFT 0.10.1.dev0
Transformers 4.41.0.dev0
Pytorch 2.2.1+cu121
Datasets 2.19.0
Tokenizers 0.19.1