Instructions to use sarathi-balakrishnan/llama-agri-b2b-intelligence-merged with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use sarathi-balakrishnan/llama-agri-b2b-intelligence-merged with Transformers:

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

pipe = pipeline("text-generation", model="sarathi-balakrishnan/llama-agri-b2b-intelligence-merged")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("sarathi-balakrishnan/llama-agri-b2b-intelligence-merged")
model = AutoModelForCausalLM.from_pretrained("sarathi-balakrishnan/llama-agri-b2b-intelligence-merged", device_map="auto")
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=256)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use sarathi-balakrishnan/llama-agri-b2b-intelligence-merged with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "sarathi-balakrishnan/llama-agri-b2b-intelligence-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": "sarathi-balakrishnan/llama-agri-b2b-intelligence-merged",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/sarathi-balakrishnan/llama-agri-b2b-intelligence-merged

SGLang

How to use sarathi-balakrishnan/llama-agri-b2b-intelligence-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 "sarathi-balakrishnan/llama-agri-b2b-intelligence-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": "sarathi-balakrishnan/llama-agri-b2b-intelligence-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 "sarathi-balakrishnan/llama-agri-b2b-intelligence-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": "sarathi-balakrishnan/llama-agri-b2b-intelligence-merged",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use sarathi-balakrishnan/llama-agri-b2b-intelligence-merged with Docker Model Runner:
```
docker model run hf.co/sarathi-balakrishnan/llama-agri-b2b-intelligence-merged
```

You need to agree to share your Hugging Face information to access this model

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

Llama 3.1 8B — Agricultural B2B Intelligence (merged weights, experiment)

Built with Llama.

Personal experiment (December 2025). This repo contains the LoRA adapter from llama-agri-b2b-intelligence merged into Llama-3.1-8B-Instruct, so it loads without PEFT. The full training details, data provenance, and limitations are on the adapter card. It has only been evaluated by loss. No task benchmark yet.

Merged vs adapter

	This repo (merged)	LoRA adapter repo
Download size	~16 GB (8.03B params in BF16)	1.34 GB adapter + the base model
Needs PEFT	No	Yes
Inference cost	Same as base Llama-3.1-8B	Small extra matmuls per layer unless merged
Outputs	Equivalent to the adapter, up to BF16 rounding from the merge	—
Good for	GGUF conversion, Ollama, plain `transformers`	Storage-limited setups, swapping adapters

Merging folds the low-rank update into each weight, W′ = W + (α/r)·B·A. After that the model runs at exactly the base model's cost. I haven't benchmarked the adapter-vs-merged speed difference, so I make no speed-up claim here.

Quick start

from transformers import AutoTokenizer, AutoModelForCausalLM
import torch

model_id = "sarathi-balakrishnan/llama-agri-b2b-intelligence-merged"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id, torch_dtype=torch.bfloat16, device_map="auto")

messages = [
    {"role": "system", "content": "You are an agricultural business intelligence analyst."},
    {"role": "user", "content": "Analyze corn market opportunities in Iowa for a seed company."},
]
text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
inputs = tokenizer(text, return_tensors="pt").to(model.device)
outputs = model.generate(**inputs, max_new_tokens=1000, temperature=0.7, do_sample=True)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[1]:], skip_special_tokens=True))

Convert to GGUF for llama.cpp / Ollama

git clone https://github.com/ggerganov/llama.cpp
cd llama.cpp && pip install -r requirements.txt

huggingface-cli download sarathi-balakrishnan/llama-agri-b2b-intelligence-merged \
    --local-dir ./agri-b2b-merged

python convert_hf_to_gguf.py ./agri-b2b-merged --outfile agri-b2b-q8.gguf --outtype q8_0

cat > Modelfile <<'MF'
FROM ./agri-b2b-q8.gguf
SYSTEM "You are an agricultural business intelligence analyst."
MF
ollama create agri-b2b -f Modelfile
ollama run agri-b2b

I haven't measured tokens/s on Apple Silicon or other local hardware for this model.

Training summary (details on the adapter card)


Method	LoRA SFT on teacher-generated responses (sequence-level distillation)
Teacher outputs	5,007 Claude Sonnet 4 + 4,993 GPT-4.1 (10,000 examples)
LoRA	r = 128, α = 256, all 7 projections, 335.5M trainable params before merging
Hardware / time	1× NVIDIA DGX Spark (GB10), ~55 h, 3 epochs
Final loss	train ~0.70, validation ~0.78
Task benchmark	Pending

About the teacher data. The training targets are outputs from third-party models. Use of this model is also subject to the terms of the providers whose models generated them (Anthropic and OpenAI), in addition to the Llama 3.1 Community License. Please review those terms for your use case.

Limitations

Not benchmarked. Trained on unverified synthetic answers, so statistics in outputs can be invented. US-only, English-only. Not for regulatory, lending, insurance, investment, or trading decisions.

License

Distributed under the Llama 3.1 Community License, subject to Meta's Acceptable Use Policy. Llama 3.1 is licensed under the Llama 3.1 Community License, Copyright © Meta Platforms, Inc. All Rights Reserved.

Citation

@misc{llama-agri-b2b-intelligence-merged-2025,
  title  = {Llama 3.1 8B Agricultural B2B Intelligence (merged): a sequence-level distillation experiment},
  author = {Sarathi Balakrishnan},
  year   = {2025},
  publisher = {Hugging Face},
  url    = {https://huggingface.co/sarathi-balakrishnan/llama-agri-b2b-intelligence-merged}
}

Downloads last month: -

Safetensors

Model size

8B params

Tensor type

BF16

Model tree for sarathi-balakrishnan/llama-agri-b2b-intelligence-merged

Base model

meta-llama/Llama-3.1-8B

Finetuned

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

Finetuned

(3307)

this model