Instructions to use Ddavidich/LOMONOSOV-ZENIT-27B-1M-INDEV with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Ddavidich/LOMONOSOV-ZENIT-27B-1M-INDEV with Transformers:

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

pipe = pipeline("text-generation", model="Ddavidich/LOMONOSOV-ZENIT-27B-1M-INDEV", trust_remote_code=True)
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, AutoModelForImageTextToText

processor = AutoProcessor.from_pretrained("Ddavidich/LOMONOSOV-ZENIT-27B-1M-INDEV", trust_remote_code=True)
model = AutoModelForImageTextToText.from_pretrained("Ddavidich/LOMONOSOV-ZENIT-27B-1M-INDEV", trust_remote_code=True, 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 Ddavidich/LOMONOSOV-ZENIT-27B-1M-INDEV with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Ddavidich/LOMONOSOV-ZENIT-27B-1M-INDEV"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Ddavidich/LOMONOSOV-ZENIT-27B-1M-INDEV",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Ddavidich/LOMONOSOV-ZENIT-27B-1M-INDEV

SGLang

How to use Ddavidich/LOMONOSOV-ZENIT-27B-1M-INDEV 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 "Ddavidich/LOMONOSOV-ZENIT-27B-1M-INDEV" \
    --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": "Ddavidich/LOMONOSOV-ZENIT-27B-1M-INDEV",
		"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 "Ddavidich/LOMONOSOV-ZENIT-27B-1M-INDEV" \
        --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": "Ddavidich/LOMONOSOV-ZENIT-27B-1M-INDEV",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Ddavidich/LOMONOSOV-ZENIT-27B-1M-INDEV with Docker Model Runner:
```
docker model run hf.co/Ddavidich/LOMONOSOV-ZENIT-27B-1M-INDEV
```

Ddavidich commited on 9 days ago

Commit

9135101

verified ·

1 Parent(s): 5d36324

Мультимодальность починена: зрительная башня переведена в FP8

Browse files

Картинка убивала движок: cuFuncSetAttribute -> CUDA_ERROR_INVALID_VALUE в
humming_gemm. Слои зрения были квантованы целочисленным pack-quantized с
группой 16, а на Blackwell её принимает только ядро humming, которое там не
запускается. Остальные отказываются: Marlin берёт -1/32/64/128, Conch -1/128,
Machete и CUTLASS требуют Hopper, AllSpark не знает sm_120.

Просто перейти на группу 128 нельзя: у linear_fc2 вход 4304 = 16 x 269.

Лестница схем, замерено на живой карте:
int g16 (было) 331 МБ — движок падает
NVFP4 g16 254 МиБ — жив, ответ мусор
FP8 поканально 445 МиБ — РАБОТАЕТ
BF16 879 МиБ — работает, подробнее

Выбран FP8: вдвое дешевле BF16 при верном ответе. Модель описывает картинку:
'LOMONOSOV ZENIT logo with a stylized emblem and text on a dark background'.

Миллион при этом сохраняется — проверено сквозным прогоном без флагов на
пустой 5090: веса 16.31 ГиБ, кэш 1 017 164 токена, выбрано окно 1 010 001,
модель отвечает.

Files changed (3) hide show

config.json +25 -58
model-vision.safetensors +2 -2
model.safetensors.index.json +0 -0

config.json CHANGED Viewed

@@ -261,74 +261,41 @@
      "zp_dtype": null
     }
    },
-   "vision_attention_w8a16_g16": {
-    "format": "pack-quantized",
-    "input_activations": null,
-    "output_activations": null,
     "targets": [
      "re:^model\\.visual\\.blocks\\.\\d+\\.attn\\.(?:qkv|proj)$",
-     "re:^visual\\.blocks\\.\\d+\\.attn\\.(?:qkv|proj)$"
     ],
     "weights": {
-     "actorder": null,
-     "block_structure": null,
-     "dynamic": false,
-     "group_size": 16,
      "num_bits": 8,
-     "observer": "memoryless_minmax",
-     "observer_kwargs": {},
-     "scale_dtype": null,
-     "strategy": "group",
      "symmetric": true,
-     "type": "int",
-     "zp_dtype": null
-    }
-   },
-   "vision_mlp_w4a16_g128": {
-    "format": "pack-quantized",
-    "input_activations": null,
-    "output_activations": null,
-    "targets": [
-     "re:^model\\.visual\\.(?:blocks\\.\\d+\\.mlp\\.linear_fc1|(?:merger|deepstack_merger_list\\.\\d+)\\.linear_fc[12])$",
-     "re:^visual\\.(?:blocks\\.\\d+\\.mlp\\.linear_fc1|(?:merger|deepstack_merger_list\\.\\d+)\\.linear_fc[12])$"
-    ],
-    "weights": {
-     "actorder": null,
-     "block_structure": null,
      "dynamic": false,
-     "group_size": 128,
-     "num_bits": 4,
-     "observer": "memoryless_minmax",
      "observer_kwargs": {},
-     "scale_dtype": null,
-     "strategy": "group",
-     "symmetric": true,
-     "type": "int",
-     "zp_dtype": null
-    }
-   },
-   "vision_mlp_w4a16_g16": {
-    "format": "pack-quantized",
-    "input_activations": null,
-    "output_activations": null,
-    "targets": [
-     "re:^model\\.visual\\.blocks\\.\\d+\\.mlp\\.linear_fc2$",
-     "re:^visual\\.blocks\\.\\d+\\.mlp\\.linear_fc2$"
-    ],
-    "weights": {
      "actorder": null,
-     "block_structure": null,
-     "dynamic": false,
-     "group_size": 16,
-     "num_bits": 4,
-     "observer": "memoryless_minmax",
-     "observer_kwargs": {},
-     "scale_dtype": null,
-     "strategy": "group",
      "symmetric": true,
-     "type": "int",
-     "zp_dtype": null
-    }
    }
   },
   "format": "mixed-precision",

      "zp_dtype": null
     }
    },
+   "vision_fp8": {
+    "format": "float-quantized",
     "targets": [
+     "re:^model\\.visual\\.(?:blocks\\.\\d+\\.mlp\\.linear_fc1|(?:merger|deepstack_merger_list\\.\\d+)\\.linear_fc[12])$",
      "re:^model\\.visual\\.blocks\\.\\d+\\.attn\\.(?:qkv|proj)$",
+     "re:^model\\.visual\\.blocks\\.\\d+\\.mlp\\.linear_fc2$",
+     "re:^visual\\.(?:blocks\\.\\d+\\.mlp\\.linear_fc1|(?:merger|deepstack_merger_list\\.\\d+)\\.linear_fc[12])$",
+     "re:^visual\\.blocks\\.\\d+\\.attn\\.(?:qkv|proj)$",
+     "re:^visual\\.blocks\\.\\d+\\.mlp\\.linear_fc2$"
     ],
     "weights": {
      "num_bits": 8,
+     "type": "float",
      "symmetric": true,
+     "strategy": "channel",
+     "group_size": null,
      "dynamic": false,
+     "observer": "minmax",
      "observer_kwargs": {},
      "actorder": null,
+     "block_structure": null
+    },
+    "input_activations": {
+     "num_bits": 8,
+     "type": "float",
      "symmetric": true,
+     "strategy": "token",
+     "group_size": null,
+     "dynamic": true,
+     "observer": null,
+     "observer_kwargs": {},
+     "actorder": null,
+     "block_structure": null
+    },
+    "output_activations": null
    }
   },
   "format": "mixed-precision",

model-vision.safetensors CHANGED Viewed

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:fc7bf6ee790c68f9a106143d040bc096793cd298ff00cad4d7691840caaf1c9a
-size 346784336

 version https://git-lfs.github.com/spec/v1
+oid sha256:06410afac298f1d1efadbbd34855e5233180d1c37faae662f3449ef7c4ff24ec
+size 466737968

model.safetensors.index.json CHANGED Viewed

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