Instructions to use RefalMachine/T-lite-it-1.0-pseudo-base with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use RefalMachine/T-lite-it-1.0-pseudo-base with Transformers:

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

pipe = pipeline("text-generation", model="RefalMachine/T-lite-it-1.0-pseudo-base")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("RefalMachine/T-lite-it-1.0-pseudo-base")
model = AutoModelForCausalLM.from_pretrained("RefalMachine/T-lite-it-1.0-pseudo-base")
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 RefalMachine/T-lite-it-1.0-pseudo-base with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "RefalMachine/T-lite-it-1.0-pseudo-base"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "RefalMachine/T-lite-it-1.0-pseudo-base",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/RefalMachine/T-lite-it-1.0-pseudo-base

SGLang

How to use RefalMachine/T-lite-it-1.0-pseudo-base 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 "RefalMachine/T-lite-it-1.0-pseudo-base" \
    --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": "RefalMachine/T-lite-it-1.0-pseudo-base",
		"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 "RefalMachine/T-lite-it-1.0-pseudo-base" \
        --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": "RefalMachine/T-lite-it-1.0-pseudo-base",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use RefalMachine/T-lite-it-1.0-pseudo-base with Docker Model Runner:
```
docker model run hf.co/RefalMachine/T-lite-it-1.0-pseudo-base
```

t-lite_part1-2_lr1e4_wsd_bs128

This model is a fine-tuned version of t-tech/T-lite-it-1.0 on an unknown dataset. It achieves the following results on the evaluation set:

Loss: 1.3980
Accuracy: 0.6669

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: 0.0001
seed: 42
distributed_type: multi-GPU
total_train_batch_size: 128
total_eval_batch_size: 128
optimizer: Adam with betas=(0.9,0.95) and epsilon=1e-05
lr_scheduler_type: warmup_stable_decay
lr_scheduler_warmup_steps: 100
num_epochs: 0.5

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
No log	0.0001	1	1.4751	0.6606
1.5071	0.0354	500	1.4113	0.6647
1.5003	0.0709	1000	1.4080	0.6649
1.4959	0.1063	1500	1.4063	0.6654
1.5019	0.1418	2000	1.4054	0.6655
1.4891	0.1772	2500	1.4047	0.6656
1.4916	0.2126	3000	1.4040	0.6657
1.496	0.2481	3500	1.4034	0.6657
1.495	0.2835	4000	1.4032	0.6657
1.4934	0.3189	4500	1.4030	0.6658
1.4849	0.3544	5000	1.4029	0.6660
1.4833	0.3898	5500	1.4024	0.6661
1.4909	0.4253	6000	1.4023	0.6661
1.4923	0.4607	6500	1.4000	0.6665
1.4965	0.4961	7000	1.3979	0.6669

Framework versions

Transformers 4.45.2
Pytorch 2.3.0a0+6ddf5cf85e.nv24.04
Datasets 2.18.0
Tokenizers 0.20.3

Downloads last month: 3

Safetensors

Model size

8B params

Tensor type

BF16

Model tree for RefalMachine/T-lite-it-1.0-pseudo-base

Base model

t-tech/T-lite-it-1.0

Finetuned

(18)

this model