Instructions to use tencent/Youtu-LLM-2B-Base with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use tencent/Youtu-LLM-2B-Base with Transformers:

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

pipe = pipeline("text-generation", model="tencent/Youtu-LLM-2B-Base")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("tencent/Youtu-LLM-2B-Base")
model = AutoModelForCausalLM.from_pretrained("tencent/Youtu-LLM-2B-Base", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use tencent/Youtu-LLM-2B-Base with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "tencent/Youtu-LLM-2B-Base"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "tencent/Youtu-LLM-2B-Base",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/tencent/Youtu-LLM-2B-Base

SGLang

How to use tencent/Youtu-LLM-2B-Base 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 "tencent/Youtu-LLM-2B-Base" \
    --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": "tencent/Youtu-LLM-2B-Base",
		"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 "tencent/Youtu-LLM-2B-Base" \
        --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": "tencent/Youtu-LLM-2B-Base",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use tencent/Youtu-LLM-2B-Base with Docker Model Runner:
```
docker model run hf.co/tencent/Youtu-LLM-2B-Base
```

remove redundant function

by Junrulu - opened Jan 9

base: refs/heads/main

←

from: refs/pr/7

Discussion Files changed

-24

Files changed (1) hide show

modeling_youtu.py +0 -24

modeling_youtu.py CHANGED Viewed

@@ -414,30 +414,6 @@ class YoutuPreTrainedModel(PreTrainedModel):
         "attentions": YoutuMLAttention,
     }
-    def init_weights(self):
-        """
-        If needed prunes and maybe initializes weights. If using a custom `PreTrainedModel`, you need to implement any
-        initialization logic in `_init_weights`.
-        """
-        # Prune heads if needed
-        if self.config.pruned_heads:
-            self.prune_heads(self.config.pruned_heads)
-        if "-init" in self.name_or_path:
-            # Initialize weights
-            self.apply(self._initialize_weights)
-            # Adjust weights of o_proj in Attention and down_proj in MLP
-            for name, module in self.named_modules():
-                if "o_proj" in name or "down_proj" in name:
-                    # For the output projection, we reinitialize the weights
-                    scaled_std = self.config.initializer_range * (1.0 / self.config.num_hidden_layers) ** 0.5
-                    module.weight.data.normal_(mean=0.0, std=scaled_std)
-            # Tie weights should be skipped when not initializing all weights
-            # since from_pretrained(...) calls tie weights anyways
-            self.tie_weights()
     def _init_weights(self, module):
         super()._init_weights(module)
         std = self.config.initializer_range

         "attentions": YoutuMLAttention,
     }
     def _init_weights(self, module):
         super()._init_weights(module)
         std = self.config.initializer_range