Instructions to use RedHatAI/Ministral-3-14B-Instruct-2512-FP8-dynamic with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use RedHatAI/Ministral-3-14B-Instruct-2512-FP8-dynamic with Transformers:

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

pipe = pipeline("text-generation", model="RedHatAI/Ministral-3-14B-Instruct-2512-FP8-dynamic")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
pipe(text=messages)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoProcessor, AutoModelForMultimodalLM

processor = AutoProcessor.from_pretrained("RedHatAI/Ministral-3-14B-Instruct-2512-FP8-dynamic")
model = AutoModelForMultimodalLM.from_pretrained("RedHatAI/Ministral-3-14B-Instruct-2512-FP8-dynamic", device_map="auto")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
inputs = processor.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=256)
print(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use RedHatAI/Ministral-3-14B-Instruct-2512-FP8-dynamic with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "RedHatAI/Ministral-3-14B-Instruct-2512-FP8-dynamic"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "RedHatAI/Ministral-3-14B-Instruct-2512-FP8-dynamic",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/RedHatAI/Ministral-3-14B-Instruct-2512-FP8-dynamic

SGLang

How to use RedHatAI/Ministral-3-14B-Instruct-2512-FP8-dynamic 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 "RedHatAI/Ministral-3-14B-Instruct-2512-FP8-dynamic" \
    --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": "RedHatAI/Ministral-3-14B-Instruct-2512-FP8-dynamic",
		"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 "RedHatAI/Ministral-3-14B-Instruct-2512-FP8-dynamic" \
        --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": "RedHatAI/Ministral-3-14B-Instruct-2512-FP8-dynamic",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use RedHatAI/Ministral-3-14B-Instruct-2512-FP8-dynamic with Docker Model Runner:
```
docker model run hf.co/RedHatAI/Ministral-3-14B-Instruct-2512-FP8-dynamic
```

ChibuUkachi commited on Feb 5

Commit

dac04dc

verified ·

1 Parent(s): 3beb36a

Upload folder using huggingface_hub

Browse files

Files changed (1) hide show

convert_ministral_hf_to_mistral.py +4 -4

convert_ministral_hf_to_mistral.py CHANGED Viewed

@@ -130,21 +130,21 @@ def convert_state_dict(hf_state_dict, config):
         if "language_model" in hf_key:
             if hf_key.endswith("q_proj.weight"):
                 tensor = permute_for_mistral_rope(tensor, text_num_attention_heads, text_query_dim, text_hidden_size)
-            elif hf_key.endswith("q_proj.weight_scale") and tensor.size(0) == text_num_attention_heads:
                 tensor = permute_for_mistral_rope(tensor, text_num_attention_heads, text_query_dim, 1)
             elif hf_key.endswith("k_proj.weight"):
                 tensor = permute_for_mistral_rope(tensor, text_num_key_value_heads, text_key_value_dim, text_hidden_size)
-            elif hf_key.endswith("k_proj.weight_scale") and tensor.size(0) == text_num_key_value_heads:
                 tensor = permute_for_mistral_rope(tensor, text_num_key_value_heads, text_key_value_dim, 1)
         if "vision_tower" in hf_key:
             if hf_key.endswith("q_proj.weight"):
                 tensor = permute_for_mistral_rope(tensor, vision_num_attention_heads, vision_query_dim, vision_hidden_size)
-            elif hf_key.endswith("q_proj.weight_scale") and tensor.size(0) == vision_num_attention_heads:
                 tensor = permute_for_mistral_rope(tensor, vision_num_attention_heads, vision_query_dim, 1)
             elif hf_key.endswith("k_proj.weight"):
                 tensor = permute_for_mistral_rope(tensor, vision_num_key_value_heads, vision_key_value_dim, vision_hidden_size)
-            elif hf_key.endswith("k_proj.weight_scale") and tensor.size(0) == vision_num_key_value_heads:
                 tensor = permute_for_mistral_rope(tensor, vision_num_key_value_heads, vision_key_value_dim, 1)
         mistral_dict[mistral_key] = tensor

         if "language_model" in hf_key:
             if hf_key.endswith("q_proj.weight"):
                 tensor = permute_for_mistral_rope(tensor, text_num_attention_heads, text_query_dim, text_hidden_size)
+            elif hf_key.endswith("q_proj.weight_scale") and tensor.size(0) > 1:
                 tensor = permute_for_mistral_rope(tensor, text_num_attention_heads, text_query_dim, 1)
             elif hf_key.endswith("k_proj.weight"):
                 tensor = permute_for_mistral_rope(tensor, text_num_key_value_heads, text_key_value_dim, text_hidden_size)
+            elif hf_key.endswith("k_proj.weight_scale") and tensor.size(0) > 1:
                 tensor = permute_for_mistral_rope(tensor, text_num_key_value_heads, text_key_value_dim, 1)
         if "vision_tower" in hf_key:
             if hf_key.endswith("q_proj.weight"):
                 tensor = permute_for_mistral_rope(tensor, vision_num_attention_heads, vision_query_dim, vision_hidden_size)
+            elif hf_key.endswith("q_proj.weight_scale") and tensor.size(0) > 1:
                 tensor = permute_for_mistral_rope(tensor, vision_num_attention_heads, vision_query_dim, 1)
             elif hf_key.endswith("k_proj.weight"):
                 tensor = permute_for_mistral_rope(tensor, vision_num_key_value_heads, vision_key_value_dim, vision_hidden_size)
+            elif hf_key.endswith("k_proj.weight_scale") and tensor.size(0) > 1:
                 tensor = permute_for_mistral_rope(tensor, vision_num_key_value_heads, vision_key_value_dim, 1)
         mistral_dict[mistral_key] = tensor