Instructions to use ThingAI/ARK-135M-Bilingual with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use ThingAI/ARK-135M-Bilingual with Transformers:

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

pipe = pipeline("text-generation", model="ThingAI/ARK-135M-Bilingual", trust_remote_code=True)
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoModelForCausalLM
model = AutoModelForCausalLM.from_pretrained("ThingAI/ARK-135M-Bilingual", trust_remote_code=True, device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use ThingAI/ARK-135M-Bilingual with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "ThingAI/ARK-135M-Bilingual"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "ThingAI/ARK-135M-Bilingual",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/ThingAI/ARK-135M-Bilingual

SGLang

How to use ThingAI/ARK-135M-Bilingual 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 "ThingAI/ARK-135M-Bilingual" \
    --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": "ThingAI/ARK-135M-Bilingual",
		"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 "ThingAI/ARK-135M-Bilingual" \
        --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": "ThingAI/ARK-135M-Bilingual",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use ThingAI/ARK-135M-Bilingual with Docker Model Runner:
```
docker model run hf.co/ThingAI/ARK-135M-Bilingual
```

ThingsAI commited on Jul 19

Commit

8fadcaf

verified ·

1 Parent(s): 709ee31

Fix RoPE inv_freq per ogni layer

Browse files

fix: recompute RoPE inv_freq buffers (were saved as NaN, causing inference failure)

Files changed (4) hide show

config.json +26 -16
generation_config.json +4 -0
model.safetensors +2 -2
tokenizer_config.json +1 -1

config.json CHANGED Viewed

@@ -1,29 +1,39 @@
 {
-  "model_type": "quark",
-  "architectures": ["QuarkForCausalLM"],
   "auto_map": {
     "AutoConfig": "configuration_quark.QuarkConfig",
     "AutoModelForCausalLM": "modeling_quark.QuarkForCausalLM"
   },
-  "vocab_size": 65537,
   "d_model": 576,
   "n_heads": 9,
   "n_kv_heads": 3,
   "n_layers": 30,
-  "d_ff": 1536,
-  "head_dim": 64,
-  "max_seq_len": 2048,
-  "rope_theta": 10000.0,
-  "rms_eps": 1e-5,
   "qkv_bias": true,
-  "dropout": 0.0,
-  "torch_dtype": "bfloat16",
-  "tie_word_embeddings": true,
   "sft_dataset": "MBZUAI/Bactrian-X (it+en)",
-  "sft_steps": 4000,
   "sft_loss": 1.9,
-  "base_pretrain": "15.7B tokens bilingual IT+EN",
   "tokenizer": "ThingAI/QuarkTokenizer",
-  "languages": ["it", "en"],
-  "special_tokens": ["<|user|>", "<|assistant|>", "<|end|>"]
-}

 {
+  "architectures": [
+    "QuarkForCausalLM"
+  ],
   "auto_map": {
     "AutoConfig": "configuration_quark.QuarkConfig",
     "AutoModelForCausalLM": "modeling_quark.QuarkForCausalLM"
   },
+  "base_pretrain": "15.7B tokens bilingual IT+EN",
+  "d_ff": 1536,
   "d_model": 576,
+  "dropout": 0.0,
+  "dtype": "float32",
+  "head_dim": 64,
+  "languages": [
+    "it",
+    "en"
+  ],
+  "max_seq_len": 2048,
+  "model_type": "quark",
   "n_heads": 9,
   "n_kv_heads": 3,
   "n_layers": 30,
   "qkv_bias": true,
+  "rms_eps": 1e-05,
+  "rope_theta": 10000.0,
   "sft_dataset": "MBZUAI/Bactrian-X (it+en)",
   "sft_loss": 1.9,
+  "sft_steps": 4000,
+  "special_tokens": [
+    "<|user|>",
+    "<|assistant|>",
+    "<|end|>"
+  ],
+  "tie_word_embeddings": true,
   "tokenizer": "ThingAI/QuarkTokenizer",
+  "transformers_version": "5.6.0",
+  "vocab_size": 65537
+}

generation_config.json ADDED Viewed

	@@ -0,0 +1,4 @@

+{
+  "_from_model_config": true,
+  "transformers_version": "5.6.0"
+}

model.safetensors CHANGED Viewed

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:8eed6899791dc95e033cd4d44ed86f08b58872a6f0bc6809d4d3f9e4a346453f
-size 363498128

 version https://git-lfs.github.com/spec/v1
+oid sha256:8b302a44dc75e61012961669ad6b11e4b4885ee57c9e89ca3debc0ce89e8dced
+size 726959640

tokenizer_config.json CHANGED Viewed

@@ -10,7 +10,7 @@
     "<|assistant|>",
     "<|end|>"
   ],
-  "is_local": true,
   "languages": [
     "en",
     "it"

     "<|assistant|>",
     "<|end|>"
   ],
+  "is_local": false,
   "languages": [
     "en",
     "it"