Tutorials

How to Create Pipeline?

Select Base Model

Retrieve base models from the Model Hub in two ways:

• Model Catalog: A repository of model sources from various providers such as DeepSeek, Gemma, Llama, and Qwen.

• Private Model: A repository for user-owned models and fine-tuned models

The Model Catalog includes the following models:

Base model	Model family	Model type	Model size	Learning stage
deepseek-ai/DeepSeek-R1-Distill-Llama-70B	DeepSeek	LLM	70B	Instruction-tuned
deepseek-ai/DeepSeek-R1-Distill-Llama-8B	DeepSeek	LLM	8B	Instruction-tuned
deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B	DeepSeek	LLM	1.5B	Instruction-tuned
deepseek-ai/DeepSeek-R1-Distill-Qwen-14B	DeepSeek	LLM	14B	Instruction-tuned
deepseek-ai/DeepSeek-R1-Distill-Qwen-32B	DeepSeek	LLM	32B	Instruction-tuned
deepseek-ai/DeepSeek-R1-Distill-Qwen-7B	DeepSeek	LLM	7B	Instruction-tuned
google/gemma-3-12b-it	Gemma	LLM	2B	Instruction-tuned
google/gemma-3-12b-pt	Gemma	LLM	2B	Pre-trained
google/gemma-3-1b-it	Gemma	LLM	1B	Instruction-tuned
google/gemma-3-1b-pt	Gemma	LLM	1B	Pre-trained
google/gemma-3-27b-it	Gemma	LLM	27B	Instruction-tuned
google/gemma-3-27b-pt	Gemma	LLM	27B	Pre-trained
google/gemma-3-4b-it	Gemma	LLM	4B	Instruction-tuned
google/medgemma-27b-text-it	Gemma	LLM (Medical)	27B	Instruction-tuned
meta-llama/Llama-3.1-70B	Llama	LLM	70B	Pre-trained
meta-llama/Llama-3.1-70B-Instruct	Llama	LLM	70B	Instruction-tuned
meta-llama/Llama-3.1-8B	Llama	LLM	8B	Pre-trained
meta-llama/Llama-3.1-8B-Instruct	Llama	LLM	8B	Instruction-tuned
meta-llama/Llama-3.2-1B	Llama	LLM	1B	Pre-trained
meta-llama/Llama-3.2-1B-Instruct	Llama	LLM	1B	Instruction-tuned
meta-llama/Llama-3.2-3B	Llama	LLM	3B	Pre-trained
meta-llama/Llama-3.2-3B-Instruct	Llama	LLM	3B	Instruction-tuned
meta-llama/Llama-3.3-70B-Instruct	Llama	LLM	70B	Instruction-tuned
mistralai/Mixtral-8x7B-Instruct-v0.1	Mistral	MoE LLM	8x7B	Instruction-tuned
mistralai/Mixtral-8x7B-v0.1	Mistral	MoE LLM	8x7B	Pre-trained
Qwen/Qwen2-0.5B	Qwen	LLM	0.5B	Pre-trained
Qwen/Qwen2-0.5B-Instruct	Qwen	LLM	0.5B	Instruction-tuned
Qwen/Qwen2-1.5B	Qwen	LLM	1.5B	Pre-trained
Qwen/Qwen2-1.5B-Instruct	Qwen	LLM	1.5B	Instruction-tuned
Qwen/Qwen2-72B	Qwen	LLM	72B	Pre-trained
Qwen/Qwen2-72B-Instruct	Qwen	LLM	72B	Instruction-tuned
Qwen/Qwen2-7B	Qwen	LLM	7B	Pre-trained
Qwen/Qwen2-7B-Instruct	Qwen	LLM	7B	Instruction-tuned
Qwen/Qwen2-VL-2B	Qwen	VLM	2B	Pre-trained
Qwen/Qwen2-VL-2B-Instruct	Qwen	VLM	2B	Instruction-tuned
Qwen/Qwen2-VL-72B	Qwen	VLM	72B	Pre-trained
Qwen/Qwen2-VL-72B-Instruct	Qwen	VLM	72B	Instruction-tuned
Qwen/Qwen2-VL-7B	Qwen	VLM	7B	Pre-trained
Qwen/Qwen2-VL-7B-Instruct	Qwen	VLM	7B	Instruction-tuned
Qwen/Qwen2.5-0.5B	Qwen	LLM	0.5B	Pre-trained
Qwen/Qwen2.5-0.5B-Instruct	Qwen	LLM	0.5B	Instruction-tuned
Qwen/Qwen2.5-1.5B	Qwen	LLM	1.5B	Pre-trained
Qwen/Qwen2.5-1.5B-Instruct	Qwen	LLM	1.5B	Instruction-tuned
Qwen/Qwen2.5-14B	Qwen	LLM	14B	Pre-trained
Qwen/Qwen2.5-14B-Instruct	Qwen	LLM	14B	Instruction-tuned
Qwen/Qwen2.5-32B	Qwen	LLM	32B	Pre-trained
Qwen/Qwen2.5-32B-Instruct	Qwen	LLM	32B	Instruction-tuned
Qwen/Qwen2.5-3B	Qwen	LLM	3B	Pre-trained
Qwen/Qwen2.5-3B-Instruct	Qwen	LLM	3B	Instruction-tuned
Qwen/Qwen2.5-72B	Qwen	LLM	72B	Pre-trained
Qwen/Qwen2.5-72B-Instruct	Qwen	LLM	72B	Instruction-tuned
Qwen/Qwen2.5-7B	Qwen	LLM	7B	Pre-trained
Qwen/Qwen2.5-7B-Instruct	Qwen	LLM	7B	Instruction-tuned
Qwen/Qwen2.5-VL-32B-Instruct	Qwen	VLM	32B	Instruction-tuned
Qwen/Qwen2.5-VL-3B-Instruct	Qwen	VLM	3B	Instruction-tuned
Qwen/Qwen2.5-VL-72B-Instruct	Qwen	VLM	72B	Instruction-tuned
Qwen/Qwen2.5-VL-7B-Instruct	Qwen	VLM	7B	Instruction-tuned
Qwen/Qwen3-0.6B	Qwen	LLM	0.6B	Pre-trained
Qwen/Qwen3-1.7B	Qwen	LLM	1.7B	Pre-trained
Qwen/Qwen3-14B	Qwen	LLM	14B	Pre-trained
Qwen/Qwen3-30B-A3B	Qwen	LLM	30B	Pre-trained
Qwen/Qwen3-32B	Qwen	LLM	32B	Pre-trained
Qwen/Qwen3-4B	Qwen	LLM	4B	Pre-trained
Qwen/Qwen3-8B	Qwen	LLM	8B	Pre-trained

The Private Model, if you want to upload your models, please contact us or follow the guide upload model through SDK.

Select Trainer

Select the appropriate trainer - which guides the model you select for training.

We offer three trainers to optimize your models:

Trainer	Definition	How it works	Best for
SFT (Supervised fine-tuning)	Foundational technique that trains your model on input-output pairs, teaching it to produce desired responses for specific inputs.	- Provide examples of correct responses to prompts to guide the model’s behavior. - Often uses human-generated “ground truth” responses to show the model how it should respond.	- Classification - Nuanced translation - Generating content in a specific format - Correcting instruction-following failures
DPO (Direct preference optimization)	Trains models to prefer certain types of responses over others by learning from comparative feedback, without requiring a separate reward model.	- Provide both correct and incorrect example responses for a prompt. - Indicate the correct response to help the model perform better.	- Summarizing text, focusing on the right things - Generating chat messages with the right tone and style
Pre-training	Initial training phase using large unlabeled data for language understanding.	- Exposes the model to vast amounts of text data to learn grammar, facts, reasoning patterns, and world knowledge. - No labeled examples required.	- Building foundational language understanding - Preparing models for downstream fine-tuning tasks

Select Volume

Choosing the right volume type ensures optimal performance and resource allocation during model fine-tuning.

We offer two volume types:

Volume	Best for	How it works
Managed volume	Your dataset is smaller than 20GiB.	- Automatically provisioned by the system. - Ideal for quick experiments and fine-tuning tasks
Dedicated network volume	Your dataset is larger than 20GiB.	- Requires manual provisioning. - You must contact the support team to request it