Instructions to use swadeshb/Llama-3.2-3B-Instruct-TRACE_GRPO-V31 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use swadeshb/Llama-3.2-3B-Instruct-TRACE_GRPO-V31 with Transformers:

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

pipe = pipeline("text-generation", model="swadeshb/Llama-3.2-3B-Instruct-TRACE_GRPO-V31")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForMultimodalLM

tokenizer = AutoTokenizer.from_pretrained("swadeshb/Llama-3.2-3B-Instruct-TRACE_GRPO-V31")
model = AutoModelForMultimodalLM.from_pretrained("swadeshb/Llama-3.2-3B-Instruct-TRACE_GRPO-V31")
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=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use swadeshb/Llama-3.2-3B-Instruct-TRACE_GRPO-V31 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "swadeshb/Llama-3.2-3B-Instruct-TRACE_GRPO-V31"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "swadeshb/Llama-3.2-3B-Instruct-TRACE_GRPO-V31",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/swadeshb/Llama-3.2-3B-Instruct-TRACE_GRPO-V31

SGLang

How to use swadeshb/Llama-3.2-3B-Instruct-TRACE_GRPO-V31 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 "swadeshb/Llama-3.2-3B-Instruct-TRACE_GRPO-V31" \
    --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": "swadeshb/Llama-3.2-3B-Instruct-TRACE_GRPO-V31",
		"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 "swadeshb/Llama-3.2-3B-Instruct-TRACE_GRPO-V31" \
        --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": "swadeshb/Llama-3.2-3B-Instruct-TRACE_GRPO-V31",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use swadeshb/Llama-3.2-3B-Instruct-TRACE_GRPO-V31 with Docker Model Runner:
```
docker model run hf.co/swadeshb/Llama-3.2-3B-Instruct-TRACE_GRPO-V31
```

swadeshb commited on Nov 11, 2025

Commit

5b5c377

verified ·

1 Parent(s): 9a699b2

Training in progress, step 100

Browse files

Files changed (4) hide show

README.md +2 -2
adapter_config.json +4 -4
adapter_model.safetensors +1 -1
training_args.bin +1 -1

README.md CHANGED Viewed

@@ -4,8 +4,8 @@ library_name: transformers
 model_name: Llama-3.2-3B-Instruct-TRACE_GRPO-V31
 tags:
 - generated_from_trainer
-- trl
 - grpo
 licence: license
 ---
@@ -27,7 +27,7 @@ print(output["generated_text"])
 ## Training procedure
-[<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="150" height="24"/>](https://wandb.ai/swadeshb-individual/huggingface/runs/7km4zsxv)
 This model was trained with GRPO, a method introduced in [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://huggingface.co/papers/2402.03300).

 model_name: Llama-3.2-3B-Instruct-TRACE_GRPO-V31
 tags:
 - generated_from_trainer
 - grpo
+- trl
 licence: license
 ---
 ## Training procedure
+[<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="150" height="24"/>](https://wandb.ai/swadeshb-individual/huggingface/runs/xawodxzk)
 This model was trained with GRPO, a method introduced in [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://huggingface.co/papers/2402.03300).

adapter_config.json CHANGED Viewed

@@ -25,13 +25,13 @@
   "rank_pattern": {},
   "revision": null,
   "target_modules": [
-    "q_proj",
     "gate_proj",
     "k_proj",
-    "up_proj",
-    "v_proj",
     "o_proj",
-    "down_proj"
   ],
   "target_parameters": null,
   "task_type": "CAUSAL_LM",

   "rank_pattern": {},
   "revision": null,
   "target_modules": [
+    "down_proj",
     "gate_proj",
+    "q_proj",
     "k_proj",
     "o_proj",
+    "v_proj",
+    "up_proj"
   ],
   "target_parameters": null,
   "task_type": "CAUSAL_LM",

adapter_model.safetensors CHANGED Viewed

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:77537aa2f5776ed74b3ed00dcc1a91eb8ecedc4572a8a3ad2b7384b849fdcc63
 size 97307544

 version https://git-lfs.github.com/spec/v1
+oid sha256:890eb184935f7aa0f601705f6e6e101c52aab31adfc3c5732558da7a06925cca
 size 97307544

training_args.bin CHANGED Viewed

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a56945cd4a599a0787cf5f4cab6f30e5882a2ea9d53ecaa0374341b169e6d1f9
 size 7377

 version https://git-lfs.github.com/spec/v1
+oid sha256:53cabfb96700124382bc40f0e7cbf792598ff955d43acc960d8f9a09d332221f
 size 7377