Instructions to use edbeeching/Qwen3-4B-Base-SFT-tr5 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use edbeeching/Qwen3-4B-Base-SFT-tr5 with Transformers:

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

pipe = pipeline("text-generation", model="edbeeching/Qwen3-4B-Base-SFT-tr5")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("edbeeching/Qwen3-4B-Base-SFT-tr5")
model = AutoModelForCausalLM.from_pretrained("edbeeching/Qwen3-4B-Base-SFT-tr5")
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]:]))

Inference
Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use edbeeching/Qwen3-4B-Base-SFT-tr5 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "edbeeching/Qwen3-4B-Base-SFT-tr5"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "edbeeching/Qwen3-4B-Base-SFT-tr5",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/edbeeching/Qwen3-4B-Base-SFT-tr5

SGLang

How to use edbeeching/Qwen3-4B-Base-SFT-tr5 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 "edbeeching/Qwen3-4B-Base-SFT-tr5" \
    --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": "edbeeching/Qwen3-4B-Base-SFT-tr5",
		"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 "edbeeching/Qwen3-4B-Base-SFT-tr5" \
        --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": "edbeeching/Qwen3-4B-Base-SFT-tr5",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use edbeeching/Qwen3-4B-Base-SFT-tr5 with Docker Model Runner:
```
docker model run hf.co/edbeeching/Qwen3-4B-Base-SFT-tr5
```

edbeeching HF Staff commited on Mar 22

Commit

55054cc

verified ·

1 Parent(s): e4c2c81

Model save

Browse files

Files changed (5) hide show

README.md +2 -0
all_results.json +5 -5
generation_config.json +1 -3
train_results.json +5 -5
trainer_state.json +0 -0

README.md CHANGED Viewed

@@ -5,7 +5,9 @@ model_name: Qwen3-4B-Base-SFT-tr5
 tags:
 - generated_from_trainer
 - trl
 - sft
 licence: license
 ---

 tags:
 - generated_from_trainer
 - trl
+- trackio:https://huggingface.co/spaces/hf-imo-colab/trackio-distillation-sft
 - sft
+- trackio
 licence: license
 ---

all_results.json CHANGED Viewed

@@ -1,9 +1,9 @@
 {
     "epoch": 4.6268656716417915,
-    "total_flos": 829952540540928.0,
-    "train_loss": 0.3511713839466533,
-    "train_runtime": 18622.9131,
     "train_samples": 4281,
-    "train_samples_per_second": 1.065,
-    "train_steps_per_second": 0.033
 }

 {
     "epoch": 4.6268656716417915,
+    "total_flos": 829202911068160.0,
+    "train_loss": 0.33651591361530364,
+    "train_runtime": 10914.7384,
     "train_samples": 4281,
+    "train_samples_per_second": 1.818,
+    "train_steps_per_second": 0.057
 }

generation_config.json CHANGED Viewed

@@ -1,8 +1,6 @@
 {
   "do_sample": false,
-  "eos_token_id": [
-    151643
-  ],
   "max_new_tokens": 2048,
   "pad_token_id": 151643,
   "transformers_version": "5.3.0.dev0"

 {
   "do_sample": false,
+  "eos_token_id": 151643,
   "max_new_tokens": 2048,
   "pad_token_id": 151643,
   "transformers_version": "5.3.0.dev0"

train_results.json CHANGED Viewed

@@ -1,9 +1,9 @@
 {
     "epoch": 4.6268656716417915,
-    "total_flos": 829952540540928.0,
-    "train_loss": 0.3511713839466533,
-    "train_runtime": 18622.9131,
     "train_samples": 4281,
-    "train_samples_per_second": 1.065,
-    "train_steps_per_second": 0.033
 }

 {
     "epoch": 4.6268656716417915,
+    "total_flos": 829202911068160.0,
+    "train_loss": 0.33651591361530364,
+    "train_runtime": 10914.7384,
     "train_samples": 4281,
+    "train_samples_per_second": 1.818,
+    "train_steps_per_second": 0.057
 }

trainer_state.json CHANGED Viewed

The diff for this file is too large to render. See raw diff