Instructions to use BucketOfFish/simplified_phi2 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use BucketOfFish/simplified_phi2 with Transformers:

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

pipe = pipeline("text-generation", model="BucketOfFish/simplified_phi2", trust_remote_code=True)

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("BucketOfFish/simplified_phi2", trust_remote_code=True, dtype="auto")

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use BucketOfFish/simplified_phi2 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "BucketOfFish/simplified_phi2"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "BucketOfFish/simplified_phi2",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/BucketOfFish/simplified_phi2

SGLang

How to use BucketOfFish/simplified_phi2 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 "BucketOfFish/simplified_phi2" \
    --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": "BucketOfFish/simplified_phi2",
		"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 "BucketOfFish/simplified_phi2" \
        --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": "BucketOfFish/simplified_phi2",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use BucketOfFish/simplified_phi2 with Docker Model Runner:
```
docker model run hf.co/BucketOfFish/simplified_phi2
```

BucketOfFish commited on Jan 6, 2024

Commit

c07c430

1 Parent(s): 455129a

Passing KV cache through iterations

Browse files

Files changed (2) hide show

phi2_model.py +3 -2
streaming_inference.py +2 -2

phi2_model.py CHANGED Viewed

@@ -35,10 +35,11 @@ class Phi2PreTrainedModel(PreTrainedModel):
     def prepare_inputs_for_generation(
         self,
         input_ids: torch.LongTensor,  # dim: (batch_size, seq_len)
-        kv_cache: KVCache | None = None,
         key_padding_mask: torch.LongTensor | torch.BoolTensor | None = None,
         **kwargs,  # has to be here
     ) -> dict[str, Any]:
         if not kv_cache:
             kv_cache = KVCache(
                 max_seqlen=self.config.initial_cos_sin_cache_len,
@@ -160,4 +161,4 @@ class Phi2ModelForCausalLM(Phi2PreTrainedModel):
             if labels is not None
             else None
         )
-        return CausalLMOutputWithPast(loss=loss, logits=logits)

     def prepare_inputs_for_generation(
         self,
         input_ids: torch.LongTensor,  # dim: (batch_size, seq_len)
+        past_key_values: KVCache | None = None,  # has to be named this
         key_padding_mask: torch.LongTensor | torch.BoolTensor | None = None,
         **kwargs,  # has to be here
     ) -> dict[str, Any]:
+        kv_cache = past_key_values
         if not kv_cache:
             kv_cache = KVCache(
                 max_seqlen=self.config.initial_cos_sin_cache_len,
             if labels is not None
             else None
         )
+        return CausalLMOutputWithPast(loss=loss, logits=logits, past_key_values=kv_cache)

streaming_inference.py CHANGED Viewed

@@ -43,11 +43,11 @@ if __name__ == "__main__":
     thread = Thread(
         target=model.generate,
         kwargs=dict(
-            tokenizer(  # returns a torch dictionary
                 "Here is an essay on sea monkeys: ",
                 return_tensors="pt",
                 return_attention_mask=False,
-                ).to(device),
             streamer=token_streamer,
             max_new_tokens=500,
             eos_token_id=tokenizer.eos_token_id,

     thread = Thread(
         target=model.generate,
         kwargs=dict(
+            inputs=tokenizer(  # returns a torch dictionary
                 "Here is an essay on sea monkeys: ",
                 return_tensors="pt",
                 return_attention_mask=False,
+            ).to(device),
             streamer=token_streamer,
             max_new_tokens=500,
             eos_token_id=tokenizer.eos_token_id,