Instructions to use togethercomputer/evo-1-131k-base with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use togethercomputer/evo-1-131k-base with Transformers:

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

pipe = pipeline("text-generation", model="togethercomputer/evo-1-131k-base", trust_remote_code=True)

# Load model directly
from transformers import AutoModelForCausalLM
model = AutoModelForCausalLM.from_pretrained("togethercomputer/evo-1-131k-base", trust_remote_code=True, dtype="auto")

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use togethercomputer/evo-1-131k-base with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "togethercomputer/evo-1-131k-base"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "togethercomputer/evo-1-131k-base",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/togethercomputer/evo-1-131k-base

SGLang

How to use togethercomputer/evo-1-131k-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 "togethercomputer/evo-1-131k-base" \
    --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/evo-1-131k-base",
		"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/evo-1-131k-base" \
        --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/evo-1-131k-base",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use togethercomputer/evo-1-131k-base with Docker Model Runner:
```
docker model run hf.co/togethercomputer/evo-1-131k-base
```

evo crash the databricks server when doing a simple inference

by qiiiibeau - opened Aug 29, 2024

Discussion

qiiiibeau

Aug 29, 2024

compute ressource: databricks azure cluster with nvidia A100

the model had no problem being loaded

code:
'''
from transformers import AutoConfig, AutoModelForCausalLM
from stripedhyena.tokenizer import CharLevelTokenizer

tokenizer = CharLevelTokenizer(512)

hf_model_name = 'togethercomputer/evo-1-131k-base'

model_config = AutoConfig.from_pretrained(
hf_model_name,
trust_remote_code=True,
revision='1.1_fix',
)

model_config.use_cache = True

model = AutoModelForCausalLM.from_pretrained(
hf_model_name,
config=model_config,
trust_remote_code=True,
revision='1.1_fix',
)

sequence = 'ACGT'

input_ids = torch.tensor(
tokenizer.tokenize(sequence),
dtype=torch.int,
).to(device).unsqueeze(0)

with torch.no_grad():
logits, _ = model(input_ids)
'''

it crashed the python kernel at the last line.
I observed exactly the same thing for stripedhyena

I was not sure if the issue was caused by configuration of my databrick ressources I also tried randomly other models of the same size e.g. "HuggingFaceH4/zephyr-7b-beta" and there were no problem making the inference. I do not know if there's any other possible incompatibility between stripedhyena/evo and databricks though.

Does anyone also encounter this problem?

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment