Instructions to use nm-testing/Inkling-Dense-54M-MTP with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use nm-testing/Inkling-Dense-54M-MTP with Transformers:

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

pipe = pipeline("image-text-to-text", model="nm-testing/Inkling-Dense-54M-MTP")
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("nm-testing/Inkling-Dense-54M-MTP")
model = AutoModelForMultimodalLM.from_pretrained("nm-testing/Inkling-Dense-54M-MTP", 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 nm-testing/Inkling-Dense-54M-MTP with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "nm-testing/Inkling-Dense-54M-MTP"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "nm-testing/Inkling-Dense-54M-MTP",
		"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/nm-testing/Inkling-Dense-54M-MTP

SGLang

How to use nm-testing/Inkling-Dense-54M-MTP 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 "nm-testing/Inkling-Dense-54M-MTP" \
    --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": "nm-testing/Inkling-Dense-54M-MTP",
		"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 "nm-testing/Inkling-Dense-54M-MTP" \
        --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": "nm-testing/Inkling-Dense-54M-MTP",
		"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 nm-testing/Inkling-Dense-54M-MTP with Docker Model Runner:
```
docker model run hf.co/nm-testing/Inkling-Dense-54M-MTP
```

Inkling-Dense-54M-MTP / fixture_provenance.json

soyrsoyr

Duplicate from soyrsoyr/Inkling-Dense-54M-MTP

1e837c7 21 days ago

Raw History Blame Contribute Delete

4.12 kB

{
  "source_commits": {
    "llm-compressor": "c892a93dad085a3a6aee7f7c72a8ac301c1e5ee9",
    "compressed-tensors": "e0b43ebab47e4ef0ecb530a499589ba9b55de16f"
  },
  "runtime": {
    "torch": {
      "distribution": "2.13.0+cu130",
      "runtime": "2.13.0+cu130",
      "path": "/usr/local/lib/python3.12/dist-packages/torch/__init__.py"
    },
    "transformers": {
      "distribution": "5.17.0",
      "runtime": "5.17.0",
      "path": "/usr/local/lib/python3.12/dist-packages/transformers/__init__.py"
    },
    "vllm": {
      "distribution": "0.30.0",
      "runtime": "0.26.0",
      "path": "/usr/local/lib/python3.12/dist-packages/vllm/__init__.py"
    },
    "accelerate": {
      "distribution": "1.15.0",
      "runtime": "1.15.0",
      "path": "/usr/local/lib/python3.12/dist-packages/accelerate/__init__.py"
    }
  },
  "runtime_note": "vLLM distribution metadata is 0.30.0 but runtime version is 0.26.0; results are specific to installed WDC build.",
  "training": {
    "eval_loss": 0.02748331055045128,
    "eval_runtime": 0.0257,
    "eval_samples_per_second": 311.39,
    "eval_steps_per_second": 77.847,
    "epoch": 100.0,
    "perplexity": 1.0278644604720464,
    "global_step": 200,
    "dataset": "create-tiny-model COPYPASTAS",
    "mtp_trained": false
  },
  "limitations": [
    "MTP layer untrained",
    "Toy perplexity evaluated on training texts",
    "Text-only, single GPU, TP=1, eager, one speculative token",
    "Does not resolve sparse Inkling MoE export issue",
    "Fully quantized MTP E2E remains failing"
  ],
  "fixture": {
    "base_model": "thinkingmachines/Inkling",
    "base_revision": "828496eeae4c243ff1a22f7f28ff83694f2f7bc9",
    "tokenizer_model": "inference-optimization/Inkling-0.92B-MTP",
    "tokenizer_revision": "22b31818016a1560cfeb47caaa3296cafa2f43df",
    "seed": 0,
    "parameters": 54136067,
    "weight_bytes": 108279142,
    "mtp_tensors": 20,
    "all_keys_in_original_checkpoint": true,
    "mtp_keys": [
      "model.mtp.layers.0.embed_norm.weight",
      "model.mtp.layers.0.hidden_norm.weight",
      "model.mtp.layers.0.input_proj.weight",
      "model.mtp.layers.0.transformer_block.attn.k_norm.weight",
      "model.mtp.layers.0.transformer_block.attn.k_sconv.weight",
      "model.mtp.layers.0.transformer_block.attn.q_norm.weight",
      "model.mtp.layers.0.transformer_block.attn.rel_logits_proj.proj",
      "model.mtp.layers.0.transformer_block.attn.v_sconv.weight",
      "model.mtp.layers.0.transformer_block.attn.wk_dv.weight",
      "model.mtp.layers.0.transformer_block.attn.wo_ud.weight",
      "model.mtp.layers.0.transformer_block.attn.wq_du.weight",
      "model.mtp.layers.0.transformer_block.attn.wr_du.weight",
      "model.mtp.layers.0.transformer_block.attn.wv_dv.weight",
      "model.mtp.layers.0.transformer_block.attn_norm.weight",
      "model.mtp.layers.0.transformer_block.attn_sconv.weight",
      "model.mtp.layers.0.transformer_block.mlp.global_scale",
      "model.mtp.layers.0.transformer_block.mlp.w13_dn.weight",
      "model.mtp.layers.0.transformer_block.mlp.w2_md.weight",
      "model.mtp.layers.0.transformer_block.mlp_norm.weight",
      "model.mtp.layers.0.transformer_block.mlp_sconv.weight"
    ],
    "backbone": "toy-trained using create-tiny-model skill",
    "mtp": "deterministically initialized, not trained"
  },
  "tested_weight_sha256": "3f37052495cee65d0c424f9dad0679b4e32f2daca26d8ae7431d3db827029ac1",
  "purpose": "Test detection, loading, quantization, and export of explicit model.mtp checkpoint layers. Separate BF16 MTP serving controls pass; fully quantized MTP serving is blocked by the tested vLLM loader.",
  "validation": {
    "bf16_mtp_serving": "passed",
    "fp8_backbone_bf16_mtp_serving": "passed",
    "fp8_mtp_export": "passed",
    "fp8_mtp_serving": "failed: vLLM Inkling MTP quant_config=None",
    "copied_mtp_serving": "passed: oneshot saved FP8 backbone, all 20 MTP tensors exactly equal to source, both prompts generated tokens, positive MTP draft counter",
    "toy_prompt": "The capital of France is -> Paris. Paris is the largest city in France..."
  }
}

	{
	"source_commits": {
	"llm-compressor": "c892a93dad085a3a6aee7f7c72a8ac301c1e5ee9",
	"compressed-tensors": "e0b43ebab47e4ef0ecb530a499589ba9b55de16f"
	},
	"runtime": {
	"torch": {
	"distribution": "2.13.0+cu130",
	"runtime": "2.13.0+cu130",
	"path": "/usr/local/lib/python3.12/dist-packages/torch/__init__.py"
	},
	"transformers": {
	"distribution": "5.17.0",
	"runtime": "5.17.0",
	"path": "/usr/local/lib/python3.12/dist-packages/transformers/__init__.py"
	},
	"vllm": {
	"distribution": "0.30.0",
	"runtime": "0.26.0",
	"path": "/usr/local/lib/python3.12/dist-packages/vllm/__init__.py"
	},
	"accelerate": {
	"distribution": "1.15.0",
	"runtime": "1.15.0",
	"path": "/usr/local/lib/python3.12/dist-packages/accelerate/__init__.py"
	}
	},
	"runtime_note": "vLLM distribution metadata is 0.30.0 but runtime version is 0.26.0; results are specific to installed WDC build.",
	"training": {
	"eval_loss": 0.02748331055045128,
	"eval_runtime": 0.0257,
	"eval_samples_per_second": 311.39,
	"eval_steps_per_second": 77.847,
	"epoch": 100.0,
	"perplexity": 1.0278644604720464,
	"global_step": 200,
	"dataset": "create-tiny-model COPYPASTAS",
	"mtp_trained": false
	},
	"limitations": [
	"MTP layer untrained",
	"Toy perplexity evaluated on training texts",
	"Text-only, single GPU, TP=1, eager, one speculative token",
	"Does not resolve sparse Inkling MoE export issue",
	"Fully quantized MTP E2E remains failing"
	],
	"fixture": {
	"base_model": "thinkingmachines/Inkling",
	"base_revision": "828496eeae4c243ff1a22f7f28ff83694f2f7bc9",
	"tokenizer_model": "inference-optimization/Inkling-0.92B-MTP",
	"tokenizer_revision": "22b31818016a1560cfeb47caaa3296cafa2f43df",
	"seed": 0,
	"parameters": 54136067,
	"weight_bytes": 108279142,
	"mtp_tensors": 20,
	"all_keys_in_original_checkpoint": true,
	"mtp_keys": [
	"model.mtp.layers.0.embed_norm.weight",
	"model.mtp.layers.0.hidden_norm.weight",
	"model.mtp.layers.0.input_proj.weight",
	"model.mtp.layers.0.transformer_block.attn.k_norm.weight",
	"model.mtp.layers.0.transformer_block.attn.k_sconv.weight",
	"model.mtp.layers.0.transformer_block.attn.q_norm.weight",
	"model.mtp.layers.0.transformer_block.attn.rel_logits_proj.proj",
	"model.mtp.layers.0.transformer_block.attn.v_sconv.weight",
	"model.mtp.layers.0.transformer_block.attn.wk_dv.weight",
	"model.mtp.layers.0.transformer_block.attn.wo_ud.weight",
	"model.mtp.layers.0.transformer_block.attn.wq_du.weight",
	"model.mtp.layers.0.transformer_block.attn.wr_du.weight",
	"model.mtp.layers.0.transformer_block.attn.wv_dv.weight",
	"model.mtp.layers.0.transformer_block.attn_norm.weight",
	"model.mtp.layers.0.transformer_block.attn_sconv.weight",
	"model.mtp.layers.0.transformer_block.mlp.global_scale",
	"model.mtp.layers.0.transformer_block.mlp.w13_dn.weight",
	"model.mtp.layers.0.transformer_block.mlp.w2_md.weight",
	"model.mtp.layers.0.transformer_block.mlp_norm.weight",
	"model.mtp.layers.0.transformer_block.mlp_sconv.weight"
	],
	"backbone": "toy-trained using create-tiny-model skill",
	"mtp": "deterministically initialized, not trained"
	},
	"tested_weight_sha256": "3f37052495cee65d0c424f9dad0679b4e32f2daca26d8ae7431d3db827029ac1",
	"purpose": "Test detection, loading, quantization, and export of explicit model.mtp checkpoint layers. Separate BF16 MTP serving controls pass; fully quantized MTP serving is blocked by the tested vLLM loader.",
	"validation": {
	"bf16_mtp_serving": "passed",
	"fp8_backbone_bf16_mtp_serving": "passed",
	"fp8_mtp_export": "passed",
	"fp8_mtp_serving": "failed: vLLM Inkling MTP quant_config=None",
	"copied_mtp_serving": "passed: oneshot saved FP8 backbone, all 20 MTP tensors exactly equal to source, both prompts generated tokens, positive MTP draft counter",
	"toy_prompt": "The capital of France is -> Paris. Paris is the largest city in France..."
	}
	}