Instructions to use NousResearch/Genstruct-7B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use NousResearch/Genstruct-7B with Transformers:

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

pipe = pipeline("text-generation", model="NousResearch/Genstruct-7B")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("NousResearch/Genstruct-7B")
model = AutoModelForCausalLM.from_pretrained("NousResearch/Genstruct-7B")
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]:]))

Inference
Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use NousResearch/Genstruct-7B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "NousResearch/Genstruct-7B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "NousResearch/Genstruct-7B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/NousResearch/Genstruct-7B

SGLang

How to use NousResearch/Genstruct-7B 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 "NousResearch/Genstruct-7B" \
    --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": "NousResearch/Genstruct-7B",
		"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 "NousResearch/Genstruct-7B" \
        --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": "NousResearch/Genstruct-7B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use NousResearch/Genstruct-7B with Docker Model Runner:
```
docker model run hf.co/NousResearch/Genstruct-7B
```

Custom Templates

by nudelbrot - opened Mar 8, 2024

Discussion

nudelbrot

Mar 8, 2024

•

edited Mar 8, 2024

HI. The model (or code) errors when not using apply_chat_template and instead feeding a manual tokenized template to the .generate() function.

How is the apply_chat_template even integrated into the model?

I'd like to change the tailing phrase "the following is..."

Same on Python 3.8, 3.11, 3.9, up to date transformers, accelerate bitsandbytes

inputs = tokenizer(prompt, return_tensors="pt").cuda()

res = tokenizer.decode(model.generate(inputs['input_ids'], max_new_tokens=512, do_sample=True, temperature=0.6)[0]).split(tokenizer.eos_token)[0]


File ~/micromamba/envs/env311_flash_attn/lib/python3.11/site-packages/transformers/generation/utils.py:1376, in GenerationMixin.generate(self, inputs, generation_config, logits_processor, stopping_criteria, prefix_allowed_tokens_fn, synced_gpus, assistant_model, streamer, negative_prompt_ids, negative_prompt_attention_mask, **kwargs)
   1368 # 3. Define model inputs
   1369 # inputs_tensor has to be defined
   1370 # model_input_name is defined if model-specific keyword input is passed
   1371 # otherwise model_input_name is None
   1372 # all model-specific keyword inputs are removed from `model_kwargs`
   1373 inputs_tensor, model_input_name, model_kwargs = self._prepare_model_inputs(
   1374     inputs, generation_config.bos_token_id, model_kwargs
   1375 )
-> 1376 batch_size = inputs_tensor.shape[0]
   1378 # 4. Define other model kwargs
   1379 model_kwargs["output_attentions"] = generation_config.output_attentions

File ~/micromamba/envs/env311_flash_attn/lib/python3.11/site-packages/transformers/tokenization_utils_base.py:268, in BatchEncoding.__getattr__(self, item)
    266     return self.data[item]
    267 except KeyError:
--> 268     raise AttributeError

AttributeError:

(no more output)

euclaise

Mar 8, 2024

•

edited Mar 8, 2024

inputs = tokenizer(prompt, return_tensors="pt").cuda()

res = tokenizer.decode(model.generate(inputs['input_ids'], max_new_tokens=512, do_sample=True, temperature=0.6)[0]).split(tokenizer.eos_token)[0]

This code is incorrect. tokenizer() returns a BatchEncoding, so I don't believe you can call .cuda() on it.

Try the following instead:

inputs = tokenizer(prompt, return_tensors="pt").input_ids.cuda()

res = tokenizer.decode(model.generate(inputs, max_new_tokens=512, do_sample=True, temperature=0.6)[0]).split(tokenizer.eos_token)[0]

Custom templates can be added as described here.

nudelbrot

Mar 9, 2024

@euclaise
you are correct, it works with the modification, thanks!

nudelbrot changed discussion status to closed Mar 9, 2024

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