Instructions to use osunlp/PreNormLoop-0.6B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use osunlp/PreNormLoop-0.6B with Transformers:

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

pipe = pipeline("text-generation", model="osunlp/PreNormLoop-0.6B", trust_remote_code=True)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoModelForCausalLM
model = AutoModelForCausalLM.from_pretrained("osunlp/PreNormLoop-0.6B", trust_remote_code=True, device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use osunlp/PreNormLoop-0.6B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "osunlp/PreNormLoop-0.6B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "osunlp/PreNormLoop-0.6B",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/osunlp/PreNormLoop-0.6B

SGLang

How to use osunlp/PreNormLoop-0.6B 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 "osunlp/PreNormLoop-0.6B" \
    --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": "osunlp/PreNormLoop-0.6B",
		"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 "osunlp/PreNormLoop-0.6B" \
        --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": "osunlp/PreNormLoop-0.6B",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use osunlp/PreNormLoop-0.6B with Docker Model Runner:
```
docker model run hf.co/osunlp/PreNormLoop-0.6B
```

ykyao commited on 1 day ago

Commit

16e089c

verified ·

1 Parent(s): 614ad44

Default to BF16 weights for inference

Browse files

Files changed (3) hide show

README.md +3 -1
config.json +3 -2
modeling_baseline.py +1 -1

README.md CHANGED Viewed

@@ -34,7 +34,7 @@ model_id = "osunlp/ARLoop-baseline-0.6B"
 device = "cuda" if torch.cuda.is_available() else "cpu"
 tokenizer = AutoTokenizer.from_pretrained(model_id, trust_remote_code=True)
 model = AutoModelForCausalLM.from_pretrained(
-    model_id, trust_remote_code=True, dtype=torch.float32,
 ).to(device).eval()
 inputs = tokenizer("The capital of France is", return_tensors="pt").to(device)
 with torch.inference_mode():
@@ -42,6 +42,8 @@ with torch.inference_mode():
 print(tokenizer.decode(output[0], skip_special_tokens=True))
 ```
 ## Architecture
 | Property | Value |

 device = "cuda" if torch.cuda.is_available() else "cpu"
 tokenizer = AutoTokenizer.from_pretrained(model_id, trust_remote_code=True)
 model = AutoModelForCausalLM.from_pretrained(
+    model_id, trust_remote_code=True, dtype=torch.bfloat16,
 ).to(device).eval()
 inputs = tokenizer("The capital of France is", return_tensors="pt").to(device)
 with torch.inference_mode():
 print(tokenizer.decode(output[0], skip_special_tokens=True))
 ```
+Weights load in BF16 for inference. The stored checkpoint retains the original FP32 weights; normalization statistics use FP32 internally.
 ## Architecture
 | Property | Value |

config.json CHANGED Viewed

@@ -7,7 +7,7 @@
     "AutoModelForCausalLM": "modeling_baseline.ARLoopBaselineForCausalLM"
   },
   "bos_token_id": 0,
-  "dtype": "float32",
   "embedding_scale": 1.0,
   "eos_token_id": 0,
   "head_dim": 96,
@@ -29,5 +29,6 @@
   "tie_word_embeddings": true,
   "transformers_version": "5.12.1",
   "use_cache": true,
-  "vocab_size": 49152
 }

     "AutoModelForCausalLM": "modeling_baseline.ARLoopBaselineForCausalLM"
   },
   "bos_token_id": 0,
+  "dtype": "bfloat16",
   "embedding_scale": 1.0,
   "eos_token_id": 0,
   "head_dim": 96,
   "tie_word_embeddings": true,
   "transformers_version": "5.12.1",
   "use_cache": true,
+  "vocab_size": 49152,
+  "torch_dtype": "bfloat16"
 }

modeling_baseline.py CHANGED Viewed

@@ -158,7 +158,7 @@ class ARLoopBaselineForCausalLM(PreTrainedModel, GenerationMixin):
             if not bool(attention_mask.all()):
                 mask = (key_indices[None, :] <= query_indices[:, None])[None, None, :, :]
                 mask = mask & attention_mask[:, None, None, :].bool()
-        # FP32 residual/value embeddings with BF16 autocast, as in training.
         with torch.autocast(device_type=input_ids.device.type, dtype=torch.bfloat16):
             h = self.transformer.wte(input_ids) * self.config.embedding_scale
             ve = self.shared_value_embedding(input_ids)

             if not bool(attention_mask.all()):
                 mask = (key_indices[None, :] <= query_indices[:, None])[None, None, :, :]
                 mask = mask & attention_mask[:, None, None, :].bool()
+        # Residual/value embeddings follow the loading dtype; Linear ops use BF16 autocast.
         with torch.autocast(device_type=input_ids.device.type, dtype=torch.bfloat16):
             h = self.transformer.wte(input_ids) * self.config.embedding_scale
             ve = self.shared_value_embedding(input_ids)