Developed by: nawazadroit
License: Apache-2.0
Base Model: unsloth/Llama-3.2-1B-Instruct
LoRA Adapter: nawazadroit/adapter_For_Vaidyslmgguf_FINAL830_1st
Quantization: Exported in 8-bit Q8_0 format
Backend Model: meta-llama/Llama-3.2-1B-Instruct (trained with float16, not 4bit)

🔧 Purpose

This is a simple tool-calling finetuned model built for testing on a specific dataset final380.json, focused on government scheme application flow in a hospital receptionist chatbot setting (Vaidya). The chatbot supports multilingual interactions (Marathi, Hindi, English) for rural and semi-urban hospital environments across Maharashtra.

🧠 Finetuning Details

| Step | Loss |

| ---- | -------- | | 45 | 0.931800 | | 46 | 0.518500 | | 47 | 0.576400 | | 48 | 0.491000 | | 49 | 1.132300 | | 50 | 0.499000 |

⏱️ Time taken: 32 minutes

📦 Completed: [50/50 32:33, Epoch 12/13]

-- 🧪 Example Inference Input Prompt:

json Copy Edit { "role": "user", "content": "Is there a special category for government employees?" } Model Output:

text Copy Edit Yes, under Category C, employees of specific organizations can apply for free or subsidized healthcare.<|eot_id|>

Dataset Used:

final380.json — focused on government healthcare scheme application and appointment booking flows.

Key Finetuning Settings:

python

from trl import SFTTrainer
from transformers import TrainingArguments, DataCollatorForSeq2Seq
from unsloth import is_bfloat16_supported

trainer = SFTTrainer(
    model = model,
    tokenizer = tokenizer,
    train_dataset = dataset,
    dataset_text_field = "text",
    max_seq_length = max_seq_length,
    data_collator = DataCollatorForSeq2Seq(tokenizer = tokenizer),
    dataset_num_proc = 1,
    packing = False,
    args = TrainingArguments(
        per_device_train_batch_size = 8,
        gradient_accumulation_steps = 16,
        warmup_steps = 10,
        num_train_epochs = 500,
        max_steps = 50,
        learning_rate = 2e-4,
        fp16 = not is_bfloat16_supported(),
        bf16 = is_bfloat16_supported(),
        logging_steps = 1,
        optim = "adamw_8bit",
        weight_decay = 0.01,
        lr_scheduler_type = "linear",
        seed = 3407,
        output_dir = "outputs",
        report_to = "none",
    ),
)

nawazadroit/vaidyaslmgguf_final3801st

Vaidya – Scheme Application Toolcalling Model (Finetuned)

🔧 Purpose

🧠 Finetuning Details

| Step | Loss |

Dataset Used:

Key Finetuning Settings:

nawazadroit/
vaidyaslmgguf_final3801st