Instructions to use emplitude/rubymelancholic with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use emplitude/rubymelancholic with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="emplitude/rubymelancholic") messages = [ {"role": "user", "content": "Who are you?"}, ] pipe(messages)# Load model directly from transformers import AutoTokenizer, AutoModelForCausalLM tokenizer = AutoTokenizer.from_pretrained("emplitude/rubymelancholic") model = AutoModelForCausalLM.from_pretrained("emplitude/rubymelancholic") 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 emplitude/rubymelancholic with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "emplitude/rubymelancholic" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "emplitude/rubymelancholic", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker
docker model run hf.co/emplitude/rubymelancholic
- SGLang
How to use emplitude/rubymelancholic 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 "emplitude/rubymelancholic" \ --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": "emplitude/rubymelancholic", "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 "emplitude/rubymelancholic" \ --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": "emplitude/rubymelancholic", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }' - Docker Model Runner
How to use emplitude/rubymelancholic with Docker Model Runner:
docker model run hf.co/emplitude/rubymelancholic
| from safetensors import safe_open | |
| from safetensors.torch import save_file | |
| import torch | |
| import hashlib | |
| import random | |
| filename = "model-00001-of-00001.safetensors" | |
| modified_model_name = "modified_model.safetensors" | |
| with safe_open(filename, framework="pt") as f: | |
| tensors = {key: f.get_tensor(key) for key in f.keys()} | |
| def introduce_noise(tensor, noise_level=1e-8): | |
| noise = torch.randn(tensor.size()) * noise_level | |
| return (tensor + noise).to(tensor.dtype) | |
| modified = 0; | |
| for key in tensors: | |
| modified +=1 | |
| tensors[key] = introduce_noise(tensors[key]) | |
| print(modified) | |
| # # modified = 0 | |
| # # modification_scale=1e-7 | |
| # # for key, tensor in tensors.items(): | |
| # # if not modified and tensor.numel() > 0: | |
| # # # Find the first non-zero element | |
| # # non_zero_indices = torch.nonzero(tensor, as_tuple=True) | |
| # # if len(non_zero_indices[0]) > 0: | |
| # # idx = tuple(index[0] for index in non_zero_indices) | |
| # # original_value = tensor[idx].item() | |
| # # # Add a tiny value | |
| # # tensor[idx] += modification_scale * abs(original_value) | |
| # # modified += 1 | |
| # # if modified == 20: | |
| # # break | |
| # # if not modified: | |
| # # print("Could not find a suitable tensor to modify.") | |
| # tensor_keys = list(tensors.keys()) | |
| # num_changes=1000 | |
| # changes_made = 0 | |
| # modification_scale=1e-7 | |
| # while changes_made < num_changes and tensor_keys: | |
| # # Randomly select a tensor | |
| # key = random.choice(tensor_keys) | |
| # tensor = tensors[key] | |
| # if tensor.numel() > 0: | |
| # # Find non-zero elements | |
| # non_zero_indices = torch.nonzero(tensor, as_tuple=False) | |
| # if len(non_zero_indices) > 0: | |
| # # Randomly select a non-zero element | |
| # idx = tuple(random.choice(non_zero_indices).tolist()) | |
| # original_value = tensor[idx].item() | |
| # # Add a small random value | |
| # modification = random.uniform(-modification_scale, modification_scale) * abs(original_value) | |
| # tensor[idx] += modification | |
| # changes_made += 1 | |
| # # Remove this tensor from the list to avoid repeated selection | |
| # tensor_keys.remove(key) | |
| # if changes_made == 0: | |
| # print("Could not find suitable tensors to modify.") | |
| # print(f"Made {changes_made} changes to the SafeTensors file.") | |
| metadata = { | |
| "format": "pt" # Adjust based on actual format needed | |
| } | |
| save_file(tensors, modified_model_name, metadata=metadata) | |
| def compute_hash(filename): | |
| hasher = hashlib.sha256() | |
| with open(filename, "rb") as f: | |
| buf = f.read() | |
| hasher.update(buf) | |
| return hasher.hexdigest() | |
| original_hash = compute_hash(filename) | |
| modified_hash = compute_hash(modified_model_name) | |
| print(f"Original Hash: {original_hash}") | |
| print(f"Modified Hash: {modified_hash}") |