Instructions to use manueldeprada/sampling_with_kvcache_hf_helpers with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use manueldeprada/sampling_with_kvcache_hf_helpers with Transformers:

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

pipe = pipeline("text-generation", model="manueldeprada/sampling_with_kvcache_hf_helpers")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("manueldeprada/sampling_with_kvcache_hf_helpers")
model = AutoModelForCausalLM.from_pretrained("manueldeprada/sampling_with_kvcache_hf_helpers", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use manueldeprada/sampling_with_kvcache_hf_helpers with vLLM:

Install from pip and serve model

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

Use Docker

docker model run hf.co/manueldeprada/sampling_with_kvcache_hf_helpers

SGLang

How to use manueldeprada/sampling_with_kvcache_hf_helpers 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 "manueldeprada/sampling_with_kvcache_hf_helpers" \
    --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": "manueldeprada/sampling_with_kvcache_hf_helpers",
		"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 "manueldeprada/sampling_with_kvcache_hf_helpers" \
        --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": "manueldeprada/sampling_with_kvcache_hf_helpers",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use manueldeprada/sampling_with_kvcache_hf_helpers with Docker Model Runner:
```
docker model run hf.co/manueldeprada/sampling_with_kvcache_hf_helpers
```

manueldeprada HF Staff commited on Jun 27, 2025

Commit

97c3e33

1 Parent(s): 47784f5

update

Browse files

Files changed (1) hide show

README.md +5 -5

README.md CHANGED Viewed

@@ -48,19 +48,19 @@ model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen2.5-0.5B-Instruct", devic
 inputs = tokenizer(["The quick brown"], return_tensors="pt").to(model.device)
 # Basic sampling
-gen_out = model.generate(**inputs, custom_generate="manueldeprada/sampling_with_kvcache", trust_remote_code=True)
 # With temperature
-gen_out = model.generate(**inputs, custom_generate="manueldeprada/sampling_with_kvcache", temperature=0.8, trust_remote_code=True)
 # With top-k
-gen_out = model.generate(**inputs, custom_generate="manueldeprada/sampling_with_kvcache", top_k=50, trust_remote_code=True)
 # With top-p (nucleus sampling)
-gen_out = model.generate(**inputs, custom_generate="manueldeprada/sampling_with_kvcache", top_p=0.9, trust_remote_code=True)
 # Greedy decoding (no sampling)
-gen_out = model.generate(**inputs, custom_generate="manueldeprada/sampling_with_kvcache", do_sample=False, trust_remote_code=True)
 # Get detailed output with probabilities
 gen_out = model.generate(

 inputs = tokenizer(["The quick brown"], return_tensors="pt").to(model.device)
 # Basic sampling
+gen_out = model.generate(**inputs, custom_generate="manueldeprada/sampling_with_kvcache_hf_helpers", trust_remote_code=True)
 # With temperature
+gen_out = model.generate(**inputs, custom_generate="manueldeprada/sampling_with_kvcache_hf_helpers", temperature=0.8, trust_remote_code=True)
 # With top-k
+gen_out = model.generate(**inputs, custom_generate="manueldeprada/sampling_with_kvcache_hf_helpers", top_k=50, trust_remote_code=True)
 # With top-p (nucleus sampling)
+gen_out = model.generate(**inputs, custom_generate="manueldeprada/sampling_with_kvcache_hf_helpers", top_p=0.9, trust_remote_code=True)
 # Greedy decoding (no sampling)
+gen_out = model.generate(**inputs, custom_generate="manueldeprada/sampling_with_kvcache_hf_helpers", do_sample=False, trust_remote_code=True)
 # Get detailed output with probabilities
 gen_out = model.generate(