Instructions to use sii-research/InnoMegrez2-Preview with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use sii-research/InnoMegrez2-Preview with Transformers:

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

pipe = pipeline("text-generation", model="sii-research/InnoMegrez2-Preview", trust_remote_code=True)
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoModelForCausalLM
model = AutoModelForCausalLM.from_pretrained("sii-research/InnoMegrez2-Preview", trust_remote_code=True, dtype="auto")

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use sii-research/InnoMegrez2-Preview with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "sii-research/InnoMegrez2-Preview"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "sii-research/InnoMegrez2-Preview",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/sii-research/InnoMegrez2-Preview

SGLang

How to use sii-research/InnoMegrez2-Preview 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 "sii-research/InnoMegrez2-Preview" \
    --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": "sii-research/InnoMegrez2-Preview",
		"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 "sii-research/InnoMegrez2-Preview" \
        --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": "sii-research/InnoMegrez2-Preview",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use sii-research/InnoMegrez2-Preview with Docker Model Runner:
```
docker model run hf.co/sii-research/InnoMegrez2-Preview
```

zyq commited on Jul 28, 2025

Commit

5aa594b

1 Parent(s): dec8751

fix rope

Browse files

Files changed (3) hide show

README.md +1 -1
README_ZH.md +1 -1
config.json +1 -1

README.md CHANGED Viewed

@@ -55,7 +55,7 @@ InnoMegrez2-Preview is a device native large language model. Megrez2 takes advan
 | **Number of Shared Experts** | 4 |
 | **Vocabulary Size** | 128,880 |
 | **Context Length** | 32K |
-| **Base Frequency of RoPE** | 1,000,000 |
 | **Attention Mechanism** | GQA |
 | **Activation Function** | SwiGLU |
 </div>

 | **Number of Shared Experts** | 4 |
 | **Vocabulary Size** | 128,880 |
 | **Context Length** | 32K |
+| **Base Frequency of RoPE** | 5,000,000 |
 | **Attention Mechanism** | GQA |
 | **Activation Function** | SwiGLU |
 </div>

README_ZH.md CHANGED Viewed

@@ -42,7 +42,7 @@ InnoMegrez2-Preview 是专为终端设备设计的大模型，兼顾MoE的精度
 | **Number of Shared Experts** | 4 |
 | **Vocabulary Size** | 128,880 |
 | **Context Length** | 32K |
-| **Base Frequency of RoPE** | 1,000,000 |
 | **Attention Mechanism** | GQA |
 | **Activation Function** | SwiGLU |
 </div>

 | **Number of Shared Experts** | 4 |
 | **Vocabulary Size** | 128,880 |
 | **Context Length** | 32K |
+| **Base Frequency of RoPE** | 5,000,000 |
 | **Attention Mechanism** | GQA |
 | **Activation Function** | SwiGLU |
 </div>

config.json CHANGED Viewed

@@ -36,7 +36,7 @@
   "pretraining_tp": 1,
   "rms_norm_eps": 1e-06,
   "rope_scaling": null,
-  "rope_theta": 1000000,
   "routed_scaling_factor": 1.0,
   "scoring_func": "softmax",
   "seq_aux": true,

   "pretraining_tp": 1,
   "rms_norm_eps": 1e-06,
   "rope_scaling": null,
+  "rope_theta": 5000000,
   "routed_scaling_factor": 1.0,
   "scoring_func": "softmax",
   "seq_aux": true,