Instructions to use yasu-oh/gemma-4-31B-it-qat-NVFP4 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use yasu-oh/gemma-4-31B-it-qat-NVFP4 with Transformers:

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

pipe = pipeline("image-text-to-text", model="yasu-oh/gemma-4-31B-it-qat-NVFP4")
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)

# Load model directly
from transformers import AutoProcessor, AutoModelForMultimodalLM

processor = AutoProcessor.from_pretrained("yasu-oh/gemma-4-31B-it-qat-NVFP4")
model = AutoModelForMultimodalLM.from_pretrained("yasu-oh/gemma-4-31B-it-qat-NVFP4", 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=40)
print(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use yasu-oh/gemma-4-31B-it-qat-NVFP4 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "yasu-oh/gemma-4-31B-it-qat-NVFP4"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "yasu-oh/gemma-4-31B-it-qat-NVFP4",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Use Docker

docker model run hf.co/yasu-oh/gemma-4-31B-it-qat-NVFP4

SGLang

How to use yasu-oh/gemma-4-31B-it-qat-NVFP4 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 "yasu-oh/gemma-4-31B-it-qat-NVFP4" \
    --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": "yasu-oh/gemma-4-31B-it-qat-NVFP4",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

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 "yasu-oh/gemma-4-31B-it-qat-NVFP4" \
        --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": "yasu-oh/gemma-4-31B-it-qat-NVFP4",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Docker Model Runner
How to use yasu-oh/gemma-4-31B-it-qat-NVFP4 with Docker Model Runner:
```
docker model run hf.co/yasu-oh/gemma-4-31B-it-qat-NVFP4
```

yasu-oh commited on 9 days ago

Commit

28413b5

1 Parent(s): f4dfe67

v3

Browse files

Files changed (5) hide show

README.md +38 -5
config.json +5 -3
model.safetensors.index.json +0 -0
recipe.yaml +36 -2
tokenizer_config.json +46 -0

README.md CHANGED Viewed

@@ -46,10 +46,44 @@ The calibration dataset consists of:
 ```yaml
 default_stage:
   default_modifiers:
     QuantizationModifier:
       targets: [Linear]
-      ignore: ['re:.*vision.*', 're:.*audio.*', lm_head, 're:.*embed.*']
-      scheme: NVFP4
       kv_cache_scheme:
         num_bits: 8
         type: float
@@ -70,13 +104,12 @@ default_stage:
 The quantization code and calibration preprocessing are available at:
-[github.com/yasu-oh/gemma-4-31b-nvfp4-quantization](https://github.com/yasu-oh/gemma-4-31b-nvfp4-quantization)
 The quantization run used:
 ```text
 transformers:        5.12.1
-compressed-tensors:  0.17.2.a20260623
-llmcompressor:       0.12.1.dev63+g57ed2be4
 ```

 ```yaml
 default_stage:
   default_modifiers:
+    IMatrixGatherer:
+      targets: [Linear]
+      ignore: ['re:.*vision.*', 're:.*audio.*', lm_head, 're:.*embed.*', 're:.*router.*']
+      weight_observer: imatrix_mse
     QuantizationModifier:
+      config_groups:
+        group_0:
+          targets: [Linear]
+          weights:
+            num_bits: 4
+            type: float
+            symmetric: true
+            group_size: 16
+            strategy: tensor_group
+            block_structure: null
+            dynamic: false
+            actorder: null
+            scale_dtype: torch.float8_e4m3fn
+            zp_dtype: null
+            observer: imatrix_mse
+            observer_kwargs: {strict: true}
+          input_activations:
+            num_bits: 4
+            type: float
+            symmetric: true
+            group_size: 16
+            strategy: tensor_group
+            block_structure: null
+            dynamic: local
+            actorder: null
+            scale_dtype: torch.float8_e4m3fn
+            zp_dtype: null
+            observer: static_minmax
+            observer_kwargs: {}
+          output_activations: null
+          format: null
       targets: [Linear]
+      ignore: ['re:.*vision.*', 're:.*audio.*', lm_head, 're:.*embed.*', 're:.*router.*']
       kv_cache_scheme:
         num_bits: 8
         type: float
 The quantization code and calibration preprocessing are available at:
+[gemma4-nvfp4-quantization](https://github.com/yasu-oh/gemma4-nvfp4-quantization)
 The quantization run used:
 ```text
 transformers:        5.12.1
+llmcompressor:       0.12.1.dev87+g9bbfd9564
 ```

config.json CHANGED Viewed

@@ -45,8 +45,10 @@
           "dynamic": false,
           "group_size": 16,
           "num_bits": 4,
-          "observer": "memoryless_minmax",
-          "observer_kwargs": {},
           "scale_dtype": "torch.float8_e4m3fn",
           "strategy": "tensor_group",
           "symmetric": true,
@@ -269,7 +271,7 @@
     "quantization_status": "compressed",
     "sparsity_config": {},
     "transform_config": {},
-    "version": "0.17.2.a20260623"
   },
   "text_config": {
     "attention_bias": false,

           "dynamic": false,
           "group_size": 16,
           "num_bits": 4,
+          "observer": "imatrix_mse",
+          "observer_kwargs": {
+            "strict": true
+          },
           "scale_dtype": "torch.float8_e4m3fn",
           "strategy": "tensor_group",
           "symmetric": true,
     "quantization_status": "compressed",
     "sparsity_config": {},
     "transform_config": {},
+    "version": "0.1.dev546+g4ed496d"
   },
   "text_config": {
     "attention_bias": false,

model.safetensors.index.json ADDED Viewed

The diff for this file is too large to render. See raw diff

recipe.yaml CHANGED Viewed

@@ -1,9 +1,43 @@
 default_stage:
   default_modifiers:
     QuantizationModifier:
       targets: [Linear]
-      ignore: ['re:.*vision.*', 're:.*audio.*', lm_head, 're:.*embed.*']
-      scheme: NVFP4
       kv_cache_scheme:
         num_bits: 8
         type: float

 default_stage:
   default_modifiers:
+    IMatrixGatherer:
+      targets: [Linear]
+      ignore: ['re:.*vision.*', 're:.*audio.*', lm_head, 're:.*embed.*', 're:.*router.*']
+      weight_observer: imatrix_mse
     QuantizationModifier:
+      config_groups:
+        group_0:
+          targets: [Linear]
+          weights:
+            num_bits: 4
+            type: float
+            symmetric: true
+            group_size: 16
+            strategy: tensor_group
+            block_structure: null
+            dynamic: false
+            actorder: null
+            scale_dtype: torch.float8_e4m3fn
+            zp_dtype: null
+            observer: imatrix_mse
+            observer_kwargs: {strict: true}
+          input_activations:
+            num_bits: 4
+            type: float
+            symmetric: true
+            group_size: 16
+            strategy: tensor_group
+            block_structure: null
+            dynamic: local
+            actorder: null
+            scale_dtype: torch.float8_e4m3fn
+            zp_dtype: null
+            observer: static_minmax
+            observer_kwargs: {}
+          output_activations: null
+          format: null
       targets: [Linear]
+      ignore: ['re:.*vision.*', 're:.*audio.*', lm_head, 're:.*embed.*', 're:.*router.*']
       kv_cache_scheme:
         num_bits: 8
         type: float

tokenizer_config.json CHANGED Viewed

@@ -85,6 +85,52 @@
     "type": "object",
     "x-regex": "(\\<\\|channel\\>thought\\n(?P<thinking>.*?)\\<channel\\|\\>)?(?P<tool_calls>\\<\\|tool_call\\>.*\\<tool_call\\|\\>)?(?P<content>(?:(?!\\<turn\\|\\>)(?!\\<\\|tool_response\\>).)+)?(?:\\<turn\\|\\>|\\<\\|tool_response\\>)?"
   },
   "soc_token": "<|channel>",
   "sot_token": "<|turn>",
   "stc_token": "<|tool_call>",

     "type": "object",
     "x-regex": "(\\<\\|channel\\>thought\\n(?P<thinking>.*?)\\<channel\\|\\>)?(?P<tool_calls>\\<\\|tool_call\\>.*\\<tool_call\\|\\>)?(?P<content>(?:(?!\\<turn\\|\\>)(?!\\<\\|tool_response\\>).)+)?(?:\\<turn\\|\\>|\\<\\|tool_response\\>)?"
   },
+  "response_template": {
+    "defaults": {
+      "role": "assistant"
+    },
+    "fields": {
+      "content": {
+        "close": [
+          "<turn|>",
+          "<|tool_response>",
+          "<eos>"
+        ],
+        "content": "text"
+      },
+      "thinking": {
+        "close": "<channel|>",
+        "content": "text",
+        "open": "<|channel>thought\n"
+      },
+      "tool_calls": {
+        "close": "<tool_call|>",
+        "content": "json",
+        "content_args": {
+          "string_delims": [
+            [
+              "<|\"|>",
+              "<|\"|>"
+            ]
+          ],
+          "unquoted_keys": true
+        },
+        "open_pattern": "<\\|tool_call>call:(?P<name>\\w+)",
+        "repeats": true,
+        "transform": {
+          "function": {
+            "arguments": "{content}",
+            "name": "{name}"
+          },
+          "type": "function"
+        }
+      }
+    },
+    "start_anchor": [
+      "<|turn>model\n",
+      "<tool_response|>"
+    ]
+  },
   "soc_token": "<|channel>",
   "sot_token": "<|turn>",
   "stc_token": "<|tool_call>",