Instructions to use ngxson/Vistral-7B-ChatML with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use ngxson/Vistral-7B-ChatML with Transformers:

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

pipe = pipeline("text-generation", model="ngxson/Vistral-7B-ChatML")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("ngxson/Vistral-7B-ChatML")
model = AutoModelForCausalLM.from_pretrained("ngxson/Vistral-7B-ChatML")
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]:]))

llama-cpp-python

How to use ngxson/Vistral-7B-ChatML with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="ngxson/Vistral-7B-ChatML",
	filename="vistral-7b-chatml-Q4_K_M.gguf",
)

llm.create_chat_completion(
	messages = [
		{
			"role": "user",
			"content": "What is the capital of France?"
		}
	]
)

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use ngxson/Vistral-7B-ChatML with llama.cpp:

Install from brew

brew install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf ngxson/Vistral-7B-ChatML:Q4_K_M
# Run inference directly in the terminal:
llama-cli -hf ngxson/Vistral-7B-ChatML:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf ngxson/Vistral-7B-ChatML:Q4_K_M
# Run inference directly in the terminal:
llama-cli -hf ngxson/Vistral-7B-ChatML:Q4_K_M

Use pre-built binary

# Download pre-built binary from:
# https://github.com/ggerganov/llama.cpp/releases
# Start a local OpenAI-compatible server with a web UI:
./llama-server -hf ngxson/Vistral-7B-ChatML:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf ngxson/Vistral-7B-ChatML:Q4_K_M

Build from source code

git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-server llama-cli
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama-server -hf ngxson/Vistral-7B-ChatML:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf ngxson/Vistral-7B-ChatML:Q4_K_M

Use Docker

docker model run hf.co/ngxson/Vistral-7B-ChatML:Q4_K_M

LM Studio
Jan

vLLM

How to use ngxson/Vistral-7B-ChatML with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "ngxson/Vistral-7B-ChatML"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "ngxson/Vistral-7B-ChatML",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/ngxson/Vistral-7B-ChatML:Q4_K_M

SGLang

How to use ngxson/Vistral-7B-ChatML 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 "ngxson/Vistral-7B-ChatML" \
    --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": "ngxson/Vistral-7B-ChatML",
		"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 "ngxson/Vistral-7B-ChatML" \
        --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": "ngxson/Vistral-7B-ChatML",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Ollama
How to use ngxson/Vistral-7B-ChatML with Ollama:
```
ollama run hf.co/ngxson/Vistral-7B-ChatML:Q4_K_M
```

Unsloth Studio

How to use ngxson/Vistral-7B-ChatML with Unsloth Studio:

Install Unsloth Studio (macOS, Linux, WSL)

curl -fsSL https://unsloth.ai/install.sh | sh
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for ngxson/Vistral-7B-ChatML to start chatting

Install Unsloth Studio (Windows)

irm https://unsloth.ai/install.ps1 | iex
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for ngxson/Vistral-7B-ChatML to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for ngxson/Vistral-7B-ChatML to start chatting

Docker Model Runner
How to use ngxson/Vistral-7B-ChatML with Docker Model Runner:
```
docker model run hf.co/ngxson/Vistral-7B-ChatML:Q4_K_M
```

Lemonade

How to use ngxson/Vistral-7B-ChatML with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull ngxson/Vistral-7B-ChatML:Q4_K_M

Run and chat with the model

lemonade run user.Vistral-7B-ChatML-Q4_K_M

List all available models

lemonade list

Vistral-7B-ChatML / finetune.py

ngxson

init

ae1a83e over 2 years ago

raw

history blame contribute delete

4.53 kB

	import torch
	from transformers import AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig, HfArgumentParser, TrainingArguments, pipeline, logging, TextStreamer
	from peft import LoraConfig, PeftModel, prepare_model_for_kbit_training, get_peft_model
	import os, torch, wandb, platform, warnings
	from datasets import load_dataset
	from trl import SFTTrainer

	hf_token = ''
	wnb_token = ''
	wnb_name = 'vistral-chatml'
	MODEL = 'Viet-Mistral/Vistral-7B-Chat'
	resume_from_checkpoint = False
	output_dir = 'vistral-chatml'
	tokenizer_path = '.'

	#######################################################
	## DATASET


	from datasets import load_dataset


	def generate_system_prompt(i):
	system_prompt = "Bạn là một trợ lí Tiếng Việt nhiệt tình và trung thực. Hãy luôn trả lời một cách hữu ích nhất có thể, đồng thời giữ an toàn."
	if i % 2 == 0:
	system_prompt += "\nCâu trả lời của bạn không nên chứa bất kỳ nội dung gây hại, phân biệt chủng tộc, phân biệt giới tính, độc hại, nguy hiểm hoặc bất hợp pháp nào. Hãy đảm bảo rằng các câu trả lời của bạn không có thiên kiến xã hội và mang tính tích cực."
	if i % 5 == 0:
	system_prompt += "\nNếu một câu hỏi không có ý nghĩa hoặc không hợp lý về mặt thông tin, hãy giải thích tại sao thay vì trả lời một điều gì đó không chính xác. Nếu bạn không biết câu trả lời cho một câu hỏi, hãy trẳ lời là bạn không biết và vui lòng không chia sẻ thông tin sai lệch."
	return system_prompt

	tokenizer = AutoTokenizer.from_pretrained(tokenizer_path)

	def tokenize_chat(input, i):
	print(generate_system_prompt(i))
	conversation = [{'role': 'system', 'content': generate_system_prompt(i)}]
	for msg in input['conversations']:
	output = {'role': 'user', 'content': msg['value']}
	if msg['from'] == 'gpt':
	output['role'] = 'assistant'
	conversation.append(output)
	formatted = tokenizer.apply_chat_template(conversation, tokenize=False)
	return tokenizer(formatted)

	sharegpt_dataset = load_dataset('bkai-foundation-models/vi-self-chat-sharegpt-format')
	train_data = sharegpt_dataset['train'].shuffle(seed=42)\
	.select(range(800))\
	.map(lambda x, i: tokenize_chat(x, i), remove_columns=["conversations"], with_indices=True)


	#######################################################
	## SETUP

	wandb.login(key=wnb_token)
	wandb.init(name=wnb_name)
	# use custom tokenizer instead of one comes from the model
	#tokenizer = AutoTokenizer.from_pretrained(
	# MODEL,
	# add_eos_token=False,
	# add_bos_token=False,
	# token=hf_token,
	#)
	bnb_config = BitsAndBytesConfig(
	load_in_4bit=True,
	bnb_4bit_quant_type="nf4",
	bnb_4bit_compute_dtype=torch.bfloat16,
	bnb_4bit_use_double_quant=True,
	)
	model = AutoModelForCausalLM.from_pretrained(
	MODEL,
	device_map="auto",
	token=hf_token,
	quantization_config=bnb_config,
	trust_remote_code=True,
	)


	#######################################################
	## LORA CONFIG

	model.gradient_checkpointing_enable()
	model = prepare_model_for_kbit_training(model)
	peft_config = LoraConfig(
	r=8,
	lora_alpha=16,
	target_modules=[
	"q_proj",
	"k_proj",
	"v_proj",
	"o_proj",
	"gate_proj",
	"up_proj",
	"down_proj",
	"lm_head",
	],
	bias="none",
	lora_dropout=0.05, # Conventional
	task_type="CAUSAL_LM",
	)
	model = get_peft_model(model, peft_config)
	model.print_trainable_parameters()

	from accelerate import Accelerator
	accelerator = Accelerator()
	model = accelerator.prepare_model(model)


	#######################################################
	## TRAIN

	from transformers import Trainer, TrainingArguments, DataCollatorForLanguageModeling
	trainer = Trainer(
	model=model,
	train_dataset=train_data,
	args=TrainingArguments(
	report_to='wandb',
	warmup_steps=1,
	per_device_train_batch_size=1,
	gradient_accumulation_steps=4,
	gradient_checkpointing=True,
	num_train_epochs=4,
	learning_rate=2.5e-5,
	logging_steps=1,
	optim="paged_adamw_8bit",
	save_strategy="steps",
	save_steps=10,
	save_total_limit=4,
	output_dir=output_dir
	),
	data_collator=DataCollatorForLanguageModeling(tokenizer, mlm=False)
	)
	model.config.use_cache = False

	trainer.train(resume_from_checkpoint=resume_from_checkpoint)