Instructions to use Sashank-810/IDC_Global_Merged with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Sashank-810/IDC_Global_Merged with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="Sashank-810/IDC_Global_Merged")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Sashank-810/IDC_Global_Merged")
model = AutoModelForCausalLM.from_pretrained("Sashank-810/IDC_Global_Merged")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use Sashank-810/IDC_Global_Merged with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Sashank-810/IDC_Global_Merged"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Sashank-810/IDC_Global_Merged",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Sashank-810/IDC_Global_Merged

SGLang

How to use Sashank-810/IDC_Global_Merged with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "Sashank-810/IDC_Global_Merged" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Sashank-810/IDC_Global_Merged",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "Sashank-810/IDC_Global_Merged" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Sashank-810/IDC_Global_Merged",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Sashank-810/IDC_Global_Merged with Docker Model Runner:
```
docker model run hf.co/Sashank-810/IDC_Global_Merged
```

LFT + IDC Math Tutor (LoRA-merged)

Summary: A math-tutor student model with an integrated IDC critic adapter merged into the base Llama-3.1-8B-Instruct (LoRA weights merged into base). Intended for math tutoring and doubt clarification.

Model Details

Base: meta-llama/Llama-3.1-8B-Instruct
Finetuned for: math tutoring + IDC-style critique/fix
Precision: FP16/BF16 compatible
Hardware: Single-GPU inference recommended

Intended Use

Educational tutoring, step-by-step math help, critique-and-fix of student answers.

Out-of-Scope

Safety-sensitive, legal, medical, or any harmful/abusive use.

How to Use (Transformers)

from transformers import AutoTokenizer, AutoModelForCausalLM

name = "Sashank-810/IDC_Global_Merged"
tok = AutoTokenizer.from_pretrained(name)
model = AutoModelForCausalLM.from_pretrained(name, torch_dtype="auto", device_map="auto")
prompt = "Explain the derivative of sin(x)."
out = model.generate(--tok(prompt, return_tensors="pt").to(model.device), max_new_tokens=128)
print(tok.decode(out[0], skip_special_tokens=True))

How to Use (vLLM)

python -m vllm.entrypoints.api_server \
  --model Sashank-810/IDC_Global_Merged \
  --dtype auto \
  --tensor-parallel-size 1

License & Responsible Use

Use responsibly for education; avoid harmful or malicious outputs.

📊 Evaluation Results (Llama 3.1-8B-Instruct Base vs Fine‑Tuned)

✅ Structured Evaluation Summary

--Total Questions:-- 2617

Base Model Performance

--Correct:-- 625
--Accuracy:-- 23.88%

Fine‑Tuned Model Performance

--Correct:-- 916
--Accuracy:-- 35.00%

🎯 Improvement

--Accuracy Gain:-- +11.12 percentage points
--Improved Answers:-- 483
--Regressed Answers:-- 192

📝 Text Generation Metrics

Base Model

--BLEU:-- 38.24 --ROUGE-1:-- 0.2947 --ROUGE-2:-- 0.0934 --ROUGE-L:-- 0.2936 --METEOR:-- 0.1633

Full Base Model Metrics

{
  "bleu": {
    "score": 38.24172039700722,
    "counts": [2214, 1378, 1110, 875],
    "totals": [3765, 2033, 1740, 1462],
    "precisions": [58.80, 67.78, 63.79, 59.85],
    "bp": 0.612276654279684,
    "sys_len": 3765,
    "ref_len": 5612
  },
  "rouge": {
    "rouge1": 0.29469964396406867,
    "rouge2": 0.09342261992242887,
    "rougeL": 0.2935582970928785,
    "rougeLsum": 0.2940696059343364
  },
  "meteor": {
    "meteor": 0.16327044830765994
  }
}

Fine‑Tuned Model

--BLEU:-- 59.31 --ROUGE-1:-- 0.4423 --ROUGE-2:-- 0.1247 --ROUGE-L:-- 0.4424 --METEOR:-- 0.2478

Full Fine‑Tuned Metrics

{
  "bleu": {
    "score": 59.31334282676538,
    "counts": [3324, 2048, 1600, 1201],
    "totals": [5734, 3124, 2659, 2219],
    "precisions": [57.97, 65.55, 60.17, 54.12],
    "bp": 1.0,
    "sys_len": 5734,
    "ref_len": 5612
  },
  "rouge": {
    "rouge1": 0.4423208144549374,
    "rouge2": 0.1247048391679649,
    "rougeL": 0.4424399985443162,
    "rougeLsum": 0.4414589284956114
  },
  "meteor": {
    "meteor": 0.24778242330127054
  }
}

Downloads last month: 3

Safetensors

Model size

8B params

Tensor type

BF16

Model tree for Sashank-810/IDC_Global_Merged

Base model

meta-llama/Llama-3.1-8B

Finetuned

meta-llama/Llama-3.1-8B-Instruct

Finetuned

Sashank-810/LFT_Final_FineTuned_Increased_Metrics

Finetuned

(1)

this model

Adapters

1 model