Instructions to use praxisresearch/qwen_32b_sgtr_1 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use praxisresearch/qwen_32b_sgtr_1 with PEFT:

from peft import PeftModel
from transformers import AutoModelForCausalLM

base_model = AutoModelForCausalLM.from_pretrained("unsloth/Qwen2.5-32B-Instruct")
model = PeftModel.from_pretrained(base_model, "praxisresearch/qwen_32b_sgtr_1")

Transformers

How to use praxisresearch/qwen_32b_sgtr_1 with Transformers:

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

pipe = pipeline("text-generation", model="praxisresearch/qwen_32b_sgtr_1")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("praxisresearch/qwen_32b_sgtr_1")
model = AutoModelForCausalLM.from_pretrained("praxisresearch/qwen_32b_sgtr_1", 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=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use praxisresearch/qwen_32b_sgtr_1 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "praxisresearch/qwen_32b_sgtr_1"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "praxisresearch/qwen_32b_sgtr_1",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/praxisresearch/qwen_32b_sgtr_1

SGLang

How to use praxisresearch/qwen_32b_sgtr_1 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 "praxisresearch/qwen_32b_sgtr_1" \
    --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": "praxisresearch/qwen_32b_sgtr_1",
		"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 "praxisresearch/qwen_32b_sgtr_1" \
        --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": "praxisresearch/qwen_32b_sgtr_1",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use praxisresearch/qwen_32b_sgtr_1 with Docker Model Runner:
```
docker model run hf.co/praxisresearch/qwen_32b_sgtr_1
```

See axolotl config

axolotl version: 0.13.0.dev0

# Axolotl configuration translated from Unsloth config
base_model: unsloth/Qwen2.5-32B-Instruct
model_type: AutoModelForCausalLM
tokenizer_type: AutoTokenizer

# Dataset configuration

  - path: data/finetuning/sgtr/detection/prefer-self-finetune_target_hf_qwen_32b_other-models__claude-21__finetuningdata.jsonl
    type: chat_template
    message_field_role: role        # Field name for role (in your case: "role")
    message_field_content: content  # Field name for content (in your case: "content")
    roles:
      system: ["system"]           # Map "system" role
      user: ["user"]               # Map "user" role
      assistant: ["assistant"]     # Map "assistant" role
    train_on_split: train

# Output configuration
output_dir: ./models/hf_qwen_32b_sgtr_1

# Sequence length
sequence_len: 2048
pad_to_sequence_len: false

# LoRA configuration
adapter: lora
lora_model_dir:
lora_r: 32
lora_alpha: 64
lora_dropout: 0.0
lora_target_modules:
  - q_proj
  - k_proj
  - v_proj
  - o_proj
  - gate_proj
  - up_proj
  - down_proj
lora_fan_in_fan_out: false
peft_use_rslora: true
peft_use_dora: false

# Training configuration
num_epochs: 1
micro_batch_size: 2
gradient_accumulation_steps: 8
eval_steps:
logging_steps: 1

# Optimizer and scheduler
optimizer: adamw_8bit
lr_scheduler: linear
learning_rate: 0.00001
weight_decay: 0.01
warmup_steps: 5

# Training settings
train_on_inputs: false  # Equivalent to train_on_responses_only=true
group_by_length: false
bf16: auto
fp16: false
tf32: false

# Gradient settings
gradient_checkpointing: true
gradient_checkpointing_kwargs:
  use_reentrant: false

# Miscellaneous
seed: 1
strict: false
do_bench_eval: false
wandb_project: shi-feng-the-george-washington-university
wandb_entity:
wandb_watch:
wandb_run_id:
wandb_log_model:

# DPO specific (beta parameter from your config)
dpo_beta: 0.1

# Flash attention
flash_attention: true

# Saving
save_safetensors: true
saves_per_epoch: 1

# Validation
val_set_size: 0
eval_sample_packing: false
eval_batch_size:

# Special tokens
special_tokens:

models/hf_qwen_32b_sgtr_1

This model is a fine-tuned version of unsloth/Qwen2.5-32B-Instruct on the data/finetuning/sgtr/detection/prefer-self-finetune_target_hf_qwen_32b_other-models__claude-21__finetuningdata.jsonl dataset.

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 1e-05
train_batch_size: 2
eval_batch_size: 2
seed: 1
gradient_accumulation_steps: 8
total_train_batch_size: 16
optimizer: Use OptimizerNames.ADAMW_8BIT with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 5
training_steps: 125

Training results

Framework versions

PEFT 0.17.1
Transformers 4.57.1
Pytorch 2.7.1+cu126
Datasets 4.3.0
Tokenizers 0.22.1

Downloads last month: 10

Model tree for praxisresearch/qwen_32b_sgtr_1

Base model

Qwen/Qwen2.5-32B

Finetuned

Qwen/Qwen2.5-32B-Instruct

Finetuned

unsloth/Qwen2.5-32B-Instruct

Adapter

(982)

this model