Instructions to use ranjanrajib/DPO-trained-Qwen2.5-Python-Coder-7B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use ranjanrajib/DPO-trained-Qwen2.5-Python-Coder-7B with Transformers:

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

pipe = pipeline("text-generation", model="ranjanrajib/DPO-trained-Qwen2.5-Python-Coder-7B")
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("ranjanrajib/DPO-trained-Qwen2.5-Python-Coder-7B")
model = AutoModelForCausalLM.from_pretrained("ranjanrajib/DPO-trained-Qwen2.5-Python-Coder-7B", 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 ranjanrajib/DPO-trained-Qwen2.5-Python-Coder-7B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "ranjanrajib/DPO-trained-Qwen2.5-Python-Coder-7B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "ranjanrajib/DPO-trained-Qwen2.5-Python-Coder-7B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/ranjanrajib/DPO-trained-Qwen2.5-Python-Coder-7B

SGLang

How to use ranjanrajib/DPO-trained-Qwen2.5-Python-Coder-7B 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 "ranjanrajib/DPO-trained-Qwen2.5-Python-Coder-7B" \
    --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": "ranjanrajib/DPO-trained-Qwen2.5-Python-Coder-7B",
		"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 "ranjanrajib/DPO-trained-Qwen2.5-Python-Coder-7B" \
        --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": "ranjanrajib/DPO-trained-Qwen2.5-Python-Coder-7B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use ranjanrajib/DPO-trained-Qwen2.5-Python-Coder-7B with Docker Model Runner:
```
docker model run hf.co/ranjanrajib/DPO-trained-Qwen2.5-Python-Coder-7B
```

ranjanrajib commited on Sep 16

Commit

e2d7367

verified ·

1 Parent(s): da2617b

Update README.md

Browse files

Files changed (1) hide show

README.md +27 -27

README.md CHANGED Viewed

@@ -87,17 +87,17 @@ was fine-tuned using:
 The resulting model demonstrates improved coding performance on the evaluated held-out benchmark:
-| Metric                  | Qwen2.5-Coder-7B-Instruct |  DPO Model |  Improvement |
 | ----------------------- | ------------------------: | ---------: | -----------: |
-| Pass@1                  |                    41.87% | ****45.22%**** | ****+3.35 pp**** |
-| Pass@5                  |                    48.87% | ****51.10%**** | ****+2.23 pp**** |
-| Pass@10                 |                    50.99% | ****53.16%**** | ****+2.17 pp**** |
-| Tests passed per answer |                    66.20% | ****69.41%**** | ****+3.21 pp**** |
 The primary result is the improvement in ****Pass@1****, meaning that a single generated solution was more likely to produce a correct solution on the evaluated programming problems.
@@ -176,7 +176,7 @@ model = PeftModel.from_pretrained(
 The base model and the merged trained model use the same VRAM. The merge changes weight values, not tensor shapes: both models have the same 7.6B parameters and architecture, so the model weights occupy the same amount of bf16 memory.
 | Component | bf16 VRAM |
-|---|---:|
 | Weights (base or merged D4) | 14.19 GiB |
 | KV cache, per token | 56 KiB |
 | Activations / workspace | ~0.5–1 GiB |
@@ -308,15 +308,15 @@ The model was evaluated against the original Qwen2.5-Coder-7B-Instruct using a c
 The same prompts, generation settings, random seeds, number of samples, and grading procedure were used for both models.
 ## Pass@k
-| Metric  | Base Model |  DPO Model |
 | ------- | ---------: | ---------: |
-| Pass@1  |     41.87% | ****45.22%**** |
-| Pass@5  |     48.87% | ****51.10%**** |
-| Pass@10 |     50.99% | ****53.16%**** |
 The improvement in Pass@1 is approximately ****3.35 percentage points****.
 ## Partial Correctness
@@ -332,17 +332,17 @@ The evaluation used the same 508 problems for both models, allowing direct compa
 For the first generated answer:
-| Base → DPO            | Problems |
 | --------------------- | -------: |
-| Correct → Correct     |      186 |
-| Correct → Incorrect   |       17 |
-| Incorrect → Correct   |       41 |
-| Incorrect → Incorrect |      264 |
 The DPO model converted 41 problems from incorrect to correct while 17 previously correct problems became incorrect.
@@ -474,15 +474,15 @@ It was constructed from programming problems drawn from:
 * handwritten programming problems
 ### Training Data
-| Dataset      | License    |
 | ------------ | ---------- |
 | BigCodeBench | Apache 2.0 |
-| MBPP         | CC BY 4.0  |
-| HumanEval    | MIT        |
 Candidate solutions were generated automatically.
@@ -531,11 +531,11 @@ The preference signal was generated from objective unit-test execution.
 The primary DPO configuration was:
 | Parameter | Value |
-|---|---:|
-| DPO beta | 0.1 |
-| Learning rate | 1e-5 |
-| Scheduler | Cosine |
-| Warmup | 5% |
 | Epochs | 2 |
 | Maximum sequence length | 1,024 |
 | Batch size | 4 |
@@ -546,7 +546,7 @@ The primary DPO configuration was:
 | Gradient checkpointing | Enabled |
 ## LoRA Configuration
 | Parameter | Value |
-|---|---|
 | Rank | 32 |
 | Alpha | 64 |
 | Dropout | 0.05 |
@@ -602,13 +602,13 @@ The DPO training objective produced a strong preference signal on the training e
 However, the validation results were substantially weaker.
-| Metric              | Training |            Validation |
 | ------------------- | -------: | --------------------: |
-| Preference accuracy |   ~84.4% |                ~45.8% |
-| Reward margin       | Positive | Approximately neutral |
 This indicates that the explicit preference signal learned during training did not generalize strongly to the validation preference examples.

 The resulting model demonstrates improved coding performance on the evaluated held-out benchmark:
+| Metric | Qwen2.5-Coder-7B-Instruct | DPO Model | Improvement |
 | ----------------------- | ------------------------: | ---------: | -----------: |
+| Pass@1 | 41.87% | ****45.22%**** | ****+3.35 pp**** |
+| Pass@5 | 48.87% | ****51.10%**** | ****+2.23 pp**** |
+| Pass@10 | 50.99% | ****53.16%**** | ****+2.17 pp**** |
+| Tests passed per answer | 66.20% | ****69.41%**** | ****+3.21 pp**** |
 The primary result is the improvement in ****Pass@1****, meaning that a single generated solution was more likely to produce a correct solution on the evaluated programming problems.
 The base model and the merged trained model use the same VRAM. The merge changes weight values, not tensor shapes: both models have the same 7.6B parameters and architecture, so the model weights occupy the same amount of bf16 memory.
 | Component | bf16 VRAM |
+| --- | ---: |
 | Weights (base or merged D4) | 14.19 GiB |
 | KV cache, per token | 56 KiB |
 | Activations / workspace | ~0.5–1 GiB |
 The same prompts, generation settings, random seeds, number of samples, and grading procedure were used for both models.
 ## Pass@k
+| Metric | Base Model | DPO Model |
 | ------- | ---------: | ---------: |
+| Pass@1 | 41.87% | ****45.22%**** |
+| Pass@5 | 48.87% | ****51.10%**** |
+| Pass@10 | 50.99% | ****53.16%**** |
 The improvement in Pass@1 is approximately ****3.35 percentage points****.
 ## Partial Correctness
 For the first generated answer:
+| Base → DPO | Problems |
 | --------------------- | -------: |
+| Correct → Correct | 186 |
+| Correct → Incorrect | 17 |
+| Incorrect → Correct | 41 |
+| Incorrect → Incorrect | 264 |
 The DPO model converted 41 problems from incorrect to correct while 17 previously correct problems became incorrect.
 * handwritten programming problems
 ### Training Data
+| Dataset | License |
 | ------------ | ---------- |
 | BigCodeBench | Apache 2.0 |
+| MBPP | CC BY 4.0 |
+| HumanEval | MIT |
 Candidate solutions were generated automatically.
 The primary DPO configuration was:
 | Parameter | Value |
+| --- | --- |
+| Rank | 32 |
+| Alpha | 64 |
+| Dropout | 0.05 |
+| Target modules | Attention and MLP projections |
 | Epochs | 2 |
 | Maximum sequence length | 1,024 |
 | Batch size | 4 |
 | Gradient checkpointing | Enabled |
 ## LoRA Configuration
 | Parameter | Value |
+| --- | --- |
 | Rank | 32 |
 | Alpha | 64 |
 | Dropout | 0.05 |
 However, the validation results were substantially weaker.
+| Metric | Training | Validation |
 | ------------------- | -------: | --------------------: |
+| Preference accuracy | ~84.4% | ~45.8% |
+| Reward margin | Positive | Approximately neutral |
 This indicates that the explicit preference signal learned during training did not generalize strongly to the validation preference examples.