Instructions to use blapuma/generative-qa-model with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use blapuma/generative-qa-model with Transformers:

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

pipe = pipeline("text-generation", model="blapuma/generative-qa-model", trust_remote_code=True)
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("blapuma/generative-qa-model", trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained("blapuma/generative-qa-model", trust_remote_code=True)
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

vLLM

How to use blapuma/generative-qa-model with vLLM:

Install from pip and serve model

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

Use Docker

docker model run hf.co/blapuma/generative-qa-model

SGLang

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

Docker Model Runner
How to use blapuma/generative-qa-model with Docker Model Runner:
```
docker model run hf.co/blapuma/generative-qa-model
```

blapuma commited on May 8, 2024

Commit

4456e69

verified ·

1 Parent(s): c82e2b8

Upload MyTestPipeline

Browse files

Files changed (1) hide show

new_task.py +5 -4

new_task.py CHANGED Viewed

@@ -16,7 +16,7 @@ class MyTestPipeline(TextGenerationPipeline):
         elif self.framework == "tf":
             in_b, input_length = tf.shape(model_inputs["input_ids"]).numpy()
-        outputs = self.model.generate(**model_inputs, **generate_kwargs, return_dict_in_generate=True, output_scores=True, max_new_tokens=10, temperature=0)
         output_ids = outputs.sequences
         out_b = output_ids.shape[0]
@@ -32,9 +32,10 @@ class MyTestPipeline(TextGenerationPipeline):
     def postprocess(self, model_outputs, **kwargs):
         guess_text = super().postprocess(model_outputs)[0]['generated_text'].split('\n')[-1].strip()
-        if guess_text[-1] == '.':
           guess_text = guess_text[:-1]
         transition_scores = self.model.compute_transition_scores(model_outputs['generated_sequence'][0], model_outputs['output_scores'], normalize_logits=True)
         log_probs = np.round(np.exp(transition_scores.cpu().numpy()), 3)[0]
         guess_prob = np.product(log_probs)
@@ -43,5 +44,5 @@ class MyTestPipeline(TextGenerationPipeline):
         confidence = max(min(confidence, 1.0), 0.0)
-        return {'guess': guess_text, 'confidence': 1.0}

         elif self.framework == "tf":
             in_b, input_length = tf.shape(model_inputs["input_ids"]).numpy()
+        outputs = self.model.generate(**model_inputs, **generate_kwargs, return_dict_in_generate=True, output_scores=True, max_new_tokens=10, do_sample=False)
         output_ids = outputs.sequences
         out_b = output_ids.shape[0]
     def postprocess(self, model_outputs, **kwargs):
         guess_text = super().postprocess(model_outputs)[0]['generated_text'].split('\n')[-1].strip()
+        # verifying that the model did generate something (protects against indexing errors)
+        if len(guess_text) > 0 and guess_text[-1] == '.':
           guess_text = guess_text[:-1]
         transition_scores = self.model.compute_transition_scores(model_outputs['generated_sequence'][0], model_outputs['output_scores'], normalize_logits=True)
         log_probs = np.round(np.exp(transition_scores.cpu().numpy()), 3)[0]
         guess_prob = np.product(log_probs)
         confidence = max(min(confidence, 1.0), 0.0)
+        return {'guess': guess_text, 'confidence': confidence}