togethercomputer
/

LLaMA-2-7B-32K

Text Generation

text-generation-inference

Model card Files Files and versions

Instructions to use togethercomputer/LLaMA-2-7B-32K with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use togethercomputer/LLaMA-2-7B-32K with Transformers:

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

pipe = pipeline("text-generation", model="togethercomputer/LLaMA-2-7B-32K")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("togethercomputer/LLaMA-2-7B-32K")
model = AutoModelForCausalLM.from_pretrained("togethercomputer/LLaMA-2-7B-32K", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

How to use togethercomputer/LLaMA-2-7B-32K with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "togethercomputer/LLaMA-2-7B-32K"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "togethercomputer/LLaMA-2-7B-32K",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/togethercomputer/LLaMA-2-7B-32K

How to use togethercomputer/LLaMA-2-7B-32K 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 "togethercomputer/LLaMA-2-7B-32K" \
    --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": "togethercomputer/LLaMA-2-7B-32K",
		"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 "togethercomputer/LLaMA-2-7B-32K" \
        --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": "togethercomputer/LLaMA-2-7B-32K",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use togethercomputer/LLaMA-2-7B-32K with Docker Model Runner:
```
docker model run hf.co/togethercomputer/LLaMA-2-7B-32K
```

Resources

View closed (5)

TemporalMesh Transformer: dynamic kNN graph attention + adaptive exit gates, 29.4 PPL at 48% compute

#41 opened 3 months ago by

research title suggestion

#40 opened almost 2 years ago by

togethercomputer/LLaMA-2-7B-32K Can be used for zeroshot classification?

#39 opened over 2 years ago by

Gonzalomoreno01

Adding `safetensors` variant of this model

#38 opened over 2 years ago by

Adding `safetensors` variant of this model

#36 opened over 2 years ago by

Adding Evaluation Results

#34 opened almost 3 years ago by

leaderboard-pr-bot

How to use on GPU

#33 opened almost 3 years ago by

Could you please explain more details about fine-tuning LLaMA-2-7B to LLaMA-2-7B-32k? Such as the fine-tuning steps and batch size. Thanks!

#32 opened almost 3 years ago by

Adding `safetensors` variant of this model

#30 opened almost 3 years ago by

RuntimeError: "addmm_impl_cpu_" not implemented for 'Half'

#29 opened almost 3 years ago by

shubhamagarwal92

Using the Accelerate API to train models on multiple GPUs

#28 opened almost 3 years ago by

Keep getting error while loading tokenizer = AutoTokenizer.from_pretrained("togethercomputer/LLaMA-2-7B-32K")

#27 opened almost 3 years ago by

[AUTOMATED] Model Memory Requirements

#26 opened almost 3 years ago by

model-sizer-bot

Installing ! pip install git+https://github.com/HazyResearch/flash-attention.git#subdirectory=csrc/rotary but flah_llama still erroring out

#25 opened almost 3 years ago by

Adding `safetensors` variant of this model

#24 opened about 3 years ago by

Quantizations for llama.cpp

#23 opened about 3 years ago by

ENDPOINT CONFIGURATION ON AWS SAGEMAKER

#21 opened about 3 years ago by

Adding `safetensors` variant of this model

#20 opened about 3 years ago by

protofile.proto: A file with this name is already in the pool

#19 opened about 3 years ago by

surya-narayanan

Is LLaMA-2-7B-32K already fine-tuned for answering questions from long text?

#18 opened about 3 years ago by

Fix RuntimeError: pad attn scores back to original query sequence length, instead of unpadded sequence length (i.e. no change).

#17 opened about 3 years ago by

How can specific information be eliminated in a LLM?

#16 opened about 3 years ago by

!pip install flash-attn --no-build-isolation

#15 opened about 3 years ago by

Instead of flash_attn it should be flash_attn_2_cuda . This is causing a deployment issue in TGI/DJL

#14 opened about 3 years ago by

RoPE scaling and max_position_embeddings

#12 opened about 3 years ago by

getting strange tokens after finetuning on Qlora

#11 opened about 3 years ago by

Training diverges when used with Llama 2 70B and 4-bit QLoRA

#10 opened about 3 years ago by

Do you have plans on making a chat model(LLama-2-7B-32k-Chat)? If so, any idea when it would come out?

#9 opened about 3 years ago by

How to training a llama-2-7B-32k from llama-2-7B?

#8 opened about 3 years ago by

Upload introduction_cn.ipynb

#7 opened about 3 years ago by

Problem with generating anything

#6 opened about 3 years ago by

Plans for 13b version?

#5 opened about 3 years ago by

GGML Version

#4 opened about 3 years ago by

Model on your API Playground

#3 opened about 3 years ago by

how to fine tune peft qlora and SFTTrainer?

#2 opened about 3 years ago by

What is the VRAM requirement of this model?

#1 opened about 3 years ago by