Instructions to use ondevicellm/tinyllama_moe_sft_ultrachat200k_v2_epochs3 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use ondevicellm/tinyllama_moe_sft_ultrachat200k_v2_epochs3 with Transformers:

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

pipe = pipeline("text-generation", model="ondevicellm/tinyllama_moe_sft_ultrachat200k_v2_epochs3")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("ondevicellm/tinyllama_moe_sft_ultrachat200k_v2_epochs3")
model = AutoModelForCausalLM.from_pretrained("ondevicellm/tinyllama_moe_sft_ultrachat200k_v2_epochs3", 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 ondevicellm/tinyllama_moe_sft_ultrachat200k_v2_epochs3 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "ondevicellm/tinyllama_moe_sft_ultrachat200k_v2_epochs3"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "ondevicellm/tinyllama_moe_sft_ultrachat200k_v2_epochs3",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/ondevicellm/tinyllama_moe_sft_ultrachat200k_v2_epochs3

SGLang

How to use ondevicellm/tinyllama_moe_sft_ultrachat200k_v2_epochs3 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 "ondevicellm/tinyllama_moe_sft_ultrachat200k_v2_epochs3" \
    --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": "ondevicellm/tinyllama_moe_sft_ultrachat200k_v2_epochs3",
		"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 "ondevicellm/tinyllama_moe_sft_ultrachat200k_v2_epochs3" \
        --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": "ondevicellm/tinyllama_moe_sft_ultrachat200k_v2_epochs3",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use ondevicellm/tinyllama_moe_sft_ultrachat200k_v2_epochs3 with Docker Model Runner:
```
docker model run hf.co/ondevicellm/tinyllama_moe_sft_ultrachat200k_v2_epochs3
```

tinyllama_moe_sft_ultrachat200k_v2_epochs3

This model is a fine-tuned version of ondevicellm/tinyllama_moe on the HuggingFaceH4/ultrachat_200k dataset. It achieves the following results on the evaluation set:

Loss: 1.1178

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: 2e-05
train_batch_size: 16
eval_batch_size: 8
seed: 42
distributed_type: multi-GPU
num_devices: 4
gradient_accumulation_steps: 2
total_train_batch_size: 128
total_eval_batch_size: 32
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: cosine
lr_scheduler_warmup_steps: 115
num_epochs: 3

Training results

Training Loss	Epoch	Step	Validation Loss
1.336	0.09	100	1.3129
1.2424	0.18	200	1.2363
1.2079	0.26	300	1.2084
1.185	0.35	400	1.1911
1.1546	0.44	500	1.1787
1.1741	0.53	600	1.1692
1.1612	0.61	700	1.1613
1.1453	0.7	800	1.1547
1.141	0.79	900	1.1489
1.1247	0.88	1000	1.1438
1.1485	0.96	1100	1.1392
1.067	1.05	1200	1.1387
1.0694	1.14	1300	1.1368
1.0814	1.23	1400	1.1341
1.0727	1.31	1500	1.1316
1.0769	1.4	1600	1.1292
1.0728	1.49	1700	1.1270
1.0558	1.58	1800	1.1247
1.0753	1.66	1900	1.1229
1.0799	1.75	2000	1.1209
1.066	1.84	2100	1.1192
1.0406	1.93	2200	1.1178
1.0193	2.01	2300	1.1222
1.0276	2.1	2400	1.1220
1.0171	2.19	2500	1.1215
1.0112	2.28	2600	1.1211
1.0087	2.37	2700	1.1207
1.0158	2.45	2800	1.1204
1.0219	2.54	2900	1.1199
1.0024	2.63	3000	1.1197
1.019	2.72	3100	1.1197
1.0135	2.8	3200	1.1194
1.0094	2.89	3300	1.1194
1.0284	2.98	3400	1.1194

Framework versions

Transformers 4.36.2
Pytorch 2.1.2+cu118
Datasets 2.14.6
Tokenizers 0.15.0

Downloads last month: 8

Safetensors

Model size

6B params

Tensor type

BF16

Model tree for ondevicellm/tinyllama_moe_sft_ultrachat200k_v2_epochs3

Base model

ondevicellm/tinyllama_moe

Finetuned

(4)

this model

Finetunes

1 model

ondevicellm
/

tinyllama_moe_sft_ultrachat200k_v2_epochs3

tinyllama_moe_sft_ultrachat200k_v2_epochs3

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

Model tree for ondevicellm/tinyllama_moe_sft_ultrachat200k_v2_epochs3

Dataset used to train ondevicellm/tinyllama_moe_sft_ultrachat200k_v2_epochs3