Text Generation
Transformers
Safetensors
Upper Grand Valley Dani
llama
genomic
text-generation-inference
Instructions to use HuggingFaceBio/Carbon-3B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use HuggingFaceBio/Carbon-3B with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="HuggingFaceBio/Carbon-3B")# Load model directly from transformers import AutoTokenizer, AutoModelForCausalLM tokenizer = AutoTokenizer.from_pretrained("HuggingFaceBio/Carbon-3B") model = AutoModelForCausalLM.from_pretrained("HuggingFaceBio/Carbon-3B") - Notebooks
- Google Colab
- Kaggle
- Local Apps
- vLLM
How to use HuggingFaceBio/Carbon-3B with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "HuggingFaceBio/Carbon-3B" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "HuggingFaceBio/Carbon-3B", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }'Use Docker
docker model run hf.co/HuggingFaceBio/Carbon-3B
- SGLang
How to use HuggingFaceBio/Carbon-3B 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 "HuggingFaceBio/Carbon-3B" \ --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": "HuggingFaceBio/Carbon-3B", "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 "HuggingFaceBio/Carbon-3B" \ --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": "HuggingFaceBio/Carbon-3B", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }' - Docker Model Runner
How to use HuggingFaceBio/Carbon-3B with Docker Model Runner:
docker model run hf.co/HuggingFaceBio/Carbon-3B
Update README.md
Browse files
README.md
CHANGED
|
@@ -245,15 +245,15 @@ Below we highlight the three short-context probes for which we report headline n
|
|
| 245 |
|
| 246 |
### Downstream tasks
|
| 247 |
|
| 248 |
-
| Category | Metric | Carbon 3B | GENERator-v2 3B | Evo2 7B (1M) |
|
| 249 |
|---|---|---|---|---|
|
| 250 |
-
| Generative | SR eukaryote
|
| 251 |
-
| Variant effect prediction | BRCA2 AUROC | **
|
| 252 |
-
| | TraitGym Mendelian AUPRC by-chrom | <u>
|
| 253 |
-
| | ClinVar coding AUROC
|
| 254 |
-
| | ClinVar non-coding AUROC
|
| 255 |
-
| Perturbation | TATA v2
|
| 256 |
-
| | SYN v2
|
| 257 |
|
| 258 |
Carbon-3B is competitive with Evo2-7B while being much faster to run.
|
| 259 |
> TODO update TATA v2 and SYN v2 scores with teh new results!
|
|
|
|
| 245 |
|
| 246 |
### Downstream tasks
|
| 247 |
|
| 248 |
+
| Category | Metric (%) | Carbon 3B | GENERator-v2 3B | Evo2 7B (1M) |
|
| 249 |
|---|---|---|---|---|
|
| 250 |
+
| Generative | SR eukaryote | **61.50** | 55.72 | <u>59.80</u> |
|
| 251 |
+
| Variant effect prediction | BRCA2 AUROC | **84.64** | 80.57 | <u>83.52</u> |
|
| 252 |
+
| | TraitGym Mendelian AUPRC by-chrom | <u>34.24</u> | 20.68 | **37.36** |
|
| 253 |
+
| | ClinVar coding AUROC, 48 kb | <u>93.30</u> | 91.98 | **93.70** |
|
| 254 |
+
| | ClinVar non-coding AUROC, 48 kb | **91.56** | <u>90.61</u> | 90.03 |
|
| 255 |
+
| Perturbation | TATA v2 | **65.94** | 49.82 | <u>63.78</u> |
|
| 256 |
+
| | SYN v2 | <u>82.78</u> | 74.08 | **84.90** |
|
| 257 |
|
| 258 |
Carbon-3B is competitive with Evo2-7B while being much faster to run.
|
| 259 |
> TODO update TATA v2 and SYN v2 scores with teh new results!
|