Instructions to use yujiepan/inkling-tiny-random with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use yujiepan/inkling-tiny-random with Transformers:

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

pipe = pipeline("image-text-to-text", model="yujiepan/inkling-tiny-random")
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("yujiepan/inkling-tiny-random")
model = AutoModelForMultimodalLM.from_pretrained("yujiepan/inkling-tiny-random", 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 yujiepan/inkling-tiny-random with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "yujiepan/inkling-tiny-random"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "yujiepan/inkling-tiny-random",
		"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/yujiepan/inkling-tiny-random

SGLang

How to use yujiepan/inkling-tiny-random 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 "yujiepan/inkling-tiny-random" \
    --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": "yujiepan/inkling-tiny-random",
		"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 "yujiepan/inkling-tiny-random" \
        --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": "yujiepan/inkling-tiny-random",
		"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 yujiepan/inkling-tiny-random with Docker Model Runner:
```
docker model run hf.co/yujiepan/inkling-tiny-random
```

yujiepan commited on Jul 21

Commit

63196aa

verified ·

1 Parent(s): af2c89d

Upload folder using huggingface_hub

Browse files

Files changed (2) hide show

README.md +14 -3
model.safetensors +2 -2

README.md CHANGED Viewed

@@ -24,7 +24,7 @@ processor = AutoProcessor.from_pretrained(model_id)
 model = AutoModelForMultimodalLM.from_pretrained(
     model_id,
     dtype=torch.bfloat16,
-    device_map="cuda",
 )
 # Synthetic multimodal inputs — no network fetch.
@@ -54,7 +54,7 @@ inputs = processor.apply_chat_template(
 ).to(model.device, dtype=model.dtype)
 input_len = inputs["input_ids"].shape[-1]
 outputs = model.generate(**inputs, max_new_tokens=16)
-print(processor.decode(outputs[0][input_len:], skip_special_tokens=False))
 ```
 ### Codes to create this repo:
@@ -103,7 +103,6 @@ config_json['text_config'].update({
     'dense_intermediate_size': 32,
     'intermediate_size': 32,
     'moe_intermediate_size': 32,
-    # 'n_routed_experts': 8,
 })
 config_json['vision_config'].update({
     'decoder_dmodel': hidden_size,
@@ -136,6 +135,14 @@ with torch.no_grad():
     for name, p in sorted(model.named_parameters()):
         torch.nn.init.normal_(p, 0, 0.2)
         print(name, p.shape, f'{p.numel() / num_params:.2%}', f'{p.numel() * p.element_size() / 1024**2:.2f}MB')
 model.save_pretrained(save_folder)
 # HF ignores `model.mtp.*` on main load; write them with original checkpoint naming.
@@ -159,6 +166,10 @@ for i in range(num_mtp_layers):
         state[name] = torch.empty(shape, dtype=torch.bfloat16)
         torch.nn.init.normal_(state[name], 0, 0.2)
         print(name, shape)
 save_file(state, str(path))
 ```

 model = AutoModelForMultimodalLM.from_pretrained(
     model_id,
     dtype=torch.bfloat16,
+    device_map="cuda" if torch.cuda.is_available() else "cpu",
 )
 # Synthetic multimodal inputs — no network fetch.
 ).to(model.device, dtype=model.dtype)
 input_len = inputs["input_ids"].shape[-1]
 outputs = model.generate(**inputs, max_new_tokens=16)
+print(processor.decode(outputs[0], skip_special_tokens=False))
 ```
 ### Codes to create this repo:
     'dense_intermediate_size': 32,
     'intermediate_size': 32,
     'moe_intermediate_size': 32,
 })
 config_json['vision_config'].update({
     'decoder_dmodel': hidden_size,
     for name, p in sorted(model.named_parameters()):
         torch.nn.init.normal_(p, 0, 0.2)
         print(name, p.shape, f'{p.numel() / num_params:.2%}', f'{p.numel() * p.element_size() / 1024**2:.2f}MB')
+# Upstream MoE gate bias / global_scale are F32; sconv stays BF16 in the checkpoint.
+for name, module in model.named_modules():
+    if hasattr(module, "e_score_correction_bias"):
+        module.e_score_correction_bias = torch.nn.Parameter(
+            module.e_score_correction_bias.detach().float()
+        )
+    if name.endswith(".mlp.gate") and hasattr(module, "global_scale"):
+        module.global_scale = torch.nn.Parameter(module.global_scale.detach().float())
 model.save_pretrained(save_folder)
 # HF ignores `model.mtp.*` on main load; write them with original checkpoint naming.
         state[name] = torch.empty(shape, dtype=torch.bfloat16)
         torch.nn.init.normal_(state[name], 0, 0.2)
         print(name, shape)
+# Keep checkpoint key dtypes aligned even if save_pretrained downcasts.
+for key, tensor in list(state.items()):
+    if key.endswith(".mlp.gate.bias") or key.endswith(".mlp.gate.global_scale"):
+        state[key] = tensor.float()
 save_file(state, str(path))
 ```

model.safetensors CHANGED Viewed

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6ca1255f3740250d220436140a50fd6603c01a31b202ccba3a773e854b5b3445
-size 7337022

 version https://git-lfs.github.com/spec/v1
+oid sha256:1b67f75e7f07e189a6dde338347e81a3616b7fed4f3236b0d34c957b69d25887
+size 7337520