Instructions to use Locutusque/TinyMistral-248M-v2 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Locutusque/TinyMistral-248M-v2 with Transformers:

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

pipe = pipeline("text-generation", model="Locutusque/TinyMistral-248M-v2")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Locutusque/TinyMistral-248M-v2")
model = AutoModelForCausalLM.from_pretrained("Locutusque/TinyMistral-248M-v2", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use Locutusque/TinyMistral-248M-v2 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Locutusque/TinyMistral-248M-v2"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Locutusque/TinyMistral-248M-v2",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/Locutusque/TinyMistral-248M-v2

SGLang

How to use Locutusque/TinyMistral-248M-v2 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 "Locutusque/TinyMistral-248M-v2" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Locutusque/TinyMistral-248M-v2",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

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 "Locutusque/TinyMistral-248M-v2" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Locutusque/TinyMistral-248M-v2",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use Locutusque/TinyMistral-248M-v2 with Docker Model Runner:
```
docker model run hf.co/Locutusque/TinyMistral-248M-v2
```

TinyMistral-248M-v2 / README.md

Locutusque

Update README.md

0f57b17 over 2 years ago

preview code

Raw

History Blame Contribute Delete

2.71 kB

metadata

license: apache-2.0
language:
  - en
pipeline_tag: text-generation
datasets:
  - Skylion007/openwebtext
  - Locutusque/TM-DATA
inference:
  parameters:
    do_sample: true
    temperature: 0.7
    top_p: 0.2
    top_k: 14
    max_new_tokens: 250
    repetition_penalty: 1.16
widget:
  - text: >-
      TITLE: Dirichlet density QUESTION [5 upvotes]: How to solve the following
      exercise: Let $q$ be prime. Show that the set of primes p for which $p
      \equiv 1\pmod q$ and $2^{(p-1)/q} \equiv 1 \pmod p$ has Dirichlet density
      $\dfrac{1}{q(q-1)}$. I want to show that $X^q-2$ (mod $p$) has a solution
      and $q$ divides $p-1$ , these two conditions are simultaneonusly satisfied
      iff p splits completely in $\Bbb{Q}(\zeta_q,2^{\frac{1}{q}})$. $\zeta_q $
      is primitive $q^{th}$ root of unity. If this is proved the I can conclude
      the result by Chebotarev density theorem. REPLY [2 votes]:
  - text: >-
      An emerging clinical approach to treat substance abuse disorders involves
      a form of cognitive-behavioral therapy whereby addicts learn to reduce
      their reactivity to drug-paired stimuli through cue-exposure or extinction
      training. It is, however,
  - text: >-
      \begin{document} \begin{frontmatter} \author{Mahouton Norbert
      Hounkonnou\corref{cor1}${}^1$}
      \cortext[cor1]{norbert.hounkonnou@cipma.uac.bj} \author{Sama
      Arjika\corref{cor2}${}^1$} \cortext[cor2]{rjksama2008@gmail.com} \author{
      Won Sang Chung\corref{cor3}${}^2$ } \cortext[cor3]{mimip4444@hanmail.net}
      \title{\bf New families of $q$ and $(q;p)-$Hermite polynomials }
      \address{${}^1$International Chair of Mathematical Physics and
      Applications \\ (ICMPA-UNESCO Chair), University of Abomey-Calavi,\\ 072
      B. P.: 50 Cotonou, Republic of Benin,\\ ${}^2$Department of Physics and
      Research Institute of Natural Science, \\ College of Natural Science, \\
      Gyeongsang National University, Jinju 660-701, Korea } \begin{abstract} In
      this paper, we construct a new family of $q-$Hermite polynomials denoted
      by $H_n(x,s|q).$ Main properties and relations are established and

Training

This model was trained on two datasets, shown in this model page.

Skylion007/openwebtext: 1,000,000 examples at a batch size of 32-4096 (1 epoch)
Locutusque/TM-DATA: All examples at a batch size of 12288 (3 epochs) Training took approximately 500 GPU hours on a single Titan V.

Metrics

You can look at the training metrics here: https://wandb.ai/locutusque/TinyMistral-V2/runs/g0rvw6wc

🔥 This model performed excellently on TruthfulQA, outperforming models more than 720x its size. These models include: mistralai/Mixtral-8x7B-v0.1, tiiuae/falcon-180B, berkeley-nest/Starling-LM-7B-alpha, upstage/SOLAR-10.7B-v1.0, and more. 🔥