Instructions to use modrill/code-think-q4b-20260908 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use modrill/code-think-q4b-20260908 with Transformers:

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

pipe = pipeline("text-generation", model="modrill/code-think-q4b-20260908")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("modrill/code-think-q4b-20260908")
model = AutoModelForCausalLM.from_pretrained("modrill/code-think-q4b-20260908", device_map="auto")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.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(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use modrill/code-think-q4b-20260908 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "modrill/code-think-q4b-20260908"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "modrill/code-think-q4b-20260908",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/modrill/code-think-q4b-20260908

SGLang

How to use modrill/code-think-q4b-20260908 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 "modrill/code-think-q4b-20260908" \
    --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": "modrill/code-think-q4b-20260908",
		"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 "modrill/code-think-q4b-20260908" \
        --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": "modrill/code-think-q4b-20260908",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use modrill/code-think-q4b-20260908 with Docker Model Runner:
```
docker model run hf.co/modrill/code-think-q4b-20260908
```

modrill commited on Sep 9

Commit

bf015da

verified ·

1 Parent(s): 16f6bfd

Add provenance/RUN_IDENTITY.json

Browse files

Files changed (1) hide show

provenance/RUN_IDENTITY.json +123 -0

provenance/RUN_IDENTITY.json ADDED Viewed

	@@ -0,0 +1,123 @@

+{
+  "schema": "RUN_IDENTITY_V1",
+  "run_id": "t30b2507-q4b-think-mix-l0.2-s0.8-v4-tail151643",
+  "corrects": "Q4B-THINK Mix Distillation 0.2:0.8 (30B-A3B-Thinking : 4B-Thinking by rows)",
+  "notes": "Mix Distillation Mix-Large (Li et al. 2025 arXiv 2502.12143): 30B-A3B-Thinking : 4B-Thinking = 0.2 : 0.8 by row/problem; one teacher trace per problem; dose target 32436894/ep ±1%, actual 1ep 32424225, 2ep 64848450. 30B source=/workspace/code-sft-runs/t30b2507-q4b-tn-v2-lr1e4-e1-20260907/data/THINK_PAIRED_V4_2EP.parquet sha=c160713fea76eae124dbbc3b5c619d3164862da83f61d91d3970ede990a12540 (2EP first half 4715 rows). 4B source=/workspace/code-sft-runs/t30b2507-q4b-tn-v2-lr1e4-e1-20260907/arms/Q4B-THINK-T4BDATA/data/joint-full-t4b_tail151643.parquet sha=d09c4d6272de7e50c9b2df0e401878864f09bda33db05a5833c43575969fa524. V4 recipe unchanged (LoRA r64/α128, lr 1e-4, both-sides B-row [151643,151667,151668], physical 2ep concat, seed 42). PI approved 2026-09-09.",
+  "model": {
+    "hf_id": "Qwen/Qwen3-4B-Base",
+    "revision": "906bfd4b4dc7f14ee4320094d8b41684abff8539",
+    "local_path": "/workspace/code-sft-infra/models/qwen3-4b-base",
+    "tied_embeddings": true
+  },
+  "data": {
+    "path": "/workspace/code-sft-runs/t30b2507-q4b-tn-v2-lr1e4-e1-20260907/arms/Q4B-THINK-MIX/data/mix_l0.2_s0.8_tail151643_2EP.parquet",
+    "rows": 8310,
+    "sha256": "6097d7bf8bc18621ed4fbc2c7f4e6a68b65186eb1d25964fd632580f97dabdfd",
+    "census": "/workspace/code-sft-runs/t30b2507-q4b-tn-v2-lr1e4-e1-20260907/arms/Q4B-THINK-MIX/data/TOKEN_CENSUS_MIX.json",
+    "physical_2ep": "concat(mix, mix) same order; teacher_source 30b/4b; tail 151643",
+    "source_1ep_path": "/workspace/code-sft-runs/t30b2507-q4b-tn-v2-lr1e4-e1-20260907/arms/Q4B-THINK-MIX/data/mix_l0.2_s0.8_tail151643.parquet",
+    "source_1ep_sha256": "3c5ae286e7e5242df3c011123edaf0727405d3b535e85d77ce6b99afe7c27e77",
+    "source_1ep_rows": 4155,
+    "teacher_30b_path": "/workspace/code-sft-runs/t30b2507-q4b-tn-v2-lr1e4-e1-20260907/data/THINK_PAIRED_V4_2EP.parquet",
+    "teacher_30b_sha256": "c160713fea76eae124dbbc3b5c619d3164862da83f61d91d3970ede990a12540",
+    "teacher_4b_path": "/workspace/code-sft-runs/t30b2507-q4b-tn-v2-lr1e4-e1-20260907/arms/Q4B-THINK-T4BDATA/data/joint-full-t4b_tail151643.parquet",
+    "teacher_4b_sha256": "d09c4d6272de7e50c9b2df0e401878864f09bda33db05a5833c43575969fa524",
+    "mix_ratio_rows": "0.2:0.8",
+    "n30": 831,
+    "n4": 3324,
+    "active_tokens_1ep": 32424225
+  },
+  "seeds": {
+    "train": 42,
+    "eval": 3407,
+    "mix": 42
+  },
+  "parameters": {
+    "epochs": 2,
+    "target_assistant_tokens": 64848450,
+    "active_tokens_1ep": 32424225,
+    "checkpoint_target_tokens": [
+      32445505,
+      48759220,
+      64848450
+    ],
+    "milestone_updates": [
+      495,
+      744,
+      990
+    ],
+    "warmup_target_tokens": 3890907,
+    "learning_rate": 0.0001,
+    "row_lr": 0.0001,
+    "optimizer": "adamw",
+    "optimizer_betas": [
+      0.9,
+      0.95
+    ],
+    "optimizer_eps": 1e-08,
+    "lora_rank": 64,
+    "lora_alpha": 128,
+    "lora_dropout": 0.0,
+    "lora_target_modules": [
+      "q_proj",
+      "k_proj",
+      "v_proj",
+      "o_proj",
+      "gate_proj",
+      "up_proj",
+      "down_proj"
+    ],
+    "lora_weight_decay": 0.1,
+    "row_weight_decay": 0.0,
+    "max_grad_norm": 1.0,
+    "tokens_per_optimizer_update": 65536,
+    "packing": false,
+    "truncation": false,
+    "repeat_examples": false,
+    "precision": "bfloat16",
+    "trainable_dtype": "float32",
+    "gradient_checkpointing": true,
+    "eos_weight": 1.0,
+    "scheduler": "cosine_by_assistant_target_token_dose",
+    "worker_lr_scale": "rstar_dual_worker_v4._lr_scale",
+    "supervised_tail": [
+      151643
+    ],
+    "trainable_token_indices": {
+      "embed_tokens": [
+        151643,
+        151667,
+        151668
+      ],
+      "lm_head": [
+        151643,
+        151667,
+        151668
+      ]
+    },
+    "untie_tied_embeddings": true,
+    "context": 32768,
+    "eval": {
+      "suite": "DEV256",
+      "seed": 3407,
+      "temperature": 0.6,
+      "top_p": 0.95,
+      "top_k": 20,
+      "ctx": 32768,
+      "stop_ids": [
+        151643,
+        151645
+      ],
+      "base_reeval_same_mode": true
+    }
+  },
+  "code": {
+    "path": "/workspace/tools/rstar_dual_worker_v4.py",
+    "commit_or_sha256": "5cbdaca65556cbe0cf9e1374cf122c22511b02781eda6fab4ca780a869562cb7",
+    "prep": "/workspace/code-sft-runs/t30b2507-q4b-tn-v2-lr1e4-e1-20260907/arms/Q4B-THINK-MIX/scripts/build_mix_distill_payload.py",
+    "prep_sha256": "2ff0f61e200d18dd55b1301f9573e923abd6fe87608220f7126cf709d4cb7288"
+  },
+  "started_at": "2026-09-09T06:13:03Z",
+  "endpoint": "2EP_ONLY @64848450 assistant tokens; no checkpoint picking",
+  "status": "RUNNING"
+}