Instructions to use March07/Qwen2.5-Coder-32B-terminal-agent-sft-allpass2040 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use March07/Qwen2.5-Coder-32B-terminal-agent-sft-allpass2040 with Transformers:

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

pipe = pipeline("text-generation", model="March07/Qwen2.5-Coder-32B-terminal-agent-sft-allpass2040")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("March07/Qwen2.5-Coder-32B-terminal-agent-sft-allpass2040")
model = AutoModelForCausalLM.from_pretrained("March07/Qwen2.5-Coder-32B-terminal-agent-sft-allpass2040", 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=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use March07/Qwen2.5-Coder-32B-terminal-agent-sft-allpass2040 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "March07/Qwen2.5-Coder-32B-terminal-agent-sft-allpass2040"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "March07/Qwen2.5-Coder-32B-terminal-agent-sft-allpass2040",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/March07/Qwen2.5-Coder-32B-terminal-agent-sft-allpass2040

SGLang

How to use March07/Qwen2.5-Coder-32B-terminal-agent-sft-allpass2040 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 "March07/Qwen2.5-Coder-32B-terminal-agent-sft-allpass2040" \
    --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": "March07/Qwen2.5-Coder-32B-terminal-agent-sft-allpass2040",
		"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 "March07/Qwen2.5-Coder-32B-terminal-agent-sft-allpass2040" \
        --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": "March07/Qwen2.5-Coder-32B-terminal-agent-sft-allpass2040",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use March07/Qwen2.5-Coder-32B-terminal-agent-sft-allpass2040 with Docker Model Runner:
```
docker model run hf.co/March07/Qwen2.5-Coder-32B-terminal-agent-sft-allpass2040
```

Qwen2.5-Coder-32B-terminal-agent-sft-allpass2040

Full-parameter SFT of Qwen/Qwen2.5-Coder-32B-Instruct on multi-turn terminal-agent trajectories.

Arm A — the all-pass baseline (highest inclusion threshold: only trajectories that passed 100% of their test cases). This is the reference point every other arm is compared against.

Ablation context

These runs study whether the gain from lowering the trajectory-inclusion threshold comes from data quality or data quantity. Lowering the threshold changes both at once (2,040 -> 3,291 trajectories, +61%), so the arms below hold N fixed and vary only composition.

arm	N	fully-passing	imperfect	imperfect %
A (all-pass baseline)	2,040	2,040	0	0%
B (size-matched mix)	2,040	1,255	785	38%
C (all-pass subset of B)	1,255	1,255	0	0%
D-dose20	2,040	1,632	408	20%
R (full, lowest threshold)	3,291	2,040	1,251	38%

This model: 2,040 trajectories, all fully-passing (0% imperfect).

Training data

Multi-turn bash-agent rollouts. Tool calls are emitted as plain text inside the assistant turn using a <tool_call>{"name": "bash", ...}</tool_call> protocol defined in the system prompt (not native function calling); tool output returns in a tool role. Loss is computed on assistant turns only.

Training procedure

Full-parameter SFT via LLaMA-Factory, DeepSpeed ZeRO-3, BF16, on 8x B200.

hyperparameter	value
learning rate	5e-6
lr scheduler	cosine, warmup ratio 0.1
weight decay	0.0
optimizer	AdamW
epochs	5
per-device batch size	1
gradient accumulation	8
effective batch size	64
max sequence length	20,000
precision	bf16
attention	sdpa

Results

Final train loss 0.2772 (from 0.9261 at the first logged step); mean train loss 0.3899 over 160 steps / 5 epochs. Runtime 2.40h.

Note: a small fraction of samples (~5%) exceed the 20,000-token cutoff and are truncated.

Loss is not comparable across arms: arms containing imperfect trajectories are fitting partly-failing rollouts, so a lower loss does not imply a better agent. These checkpoints must be compared by task success rate, not by training loss.

Framework versions

Transformers 5.8.0
PyTorch 2.11.0+cu130
DeepSpeed 0.18.9
Datasets 4.0.0

Usage

from transformers import AutoModelForCausalLM, AutoTokenizer

model_id = "March07/Qwen2.5-Coder-32B-terminal-agent-sft-allpass2040"
tok = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id, dtype="bfloat16", device_map="auto")

messages = [
    {"role": "system", "content": "You are an expert technical assistant with access to bash tools."},
    {"role": "user", "content": "List the files in /workspace."},
]
inputs = tok.apply_chat_template(messages, add_generation_prompt=True, return_tensors="pt").to(model.device)
out = model.generate(inputs, max_new_tokens=512)
print(tok.decode(out[0][inputs.shape[-1]:], skip_special_tokens=True))

Downloads last month: 6

Safetensors

Model size

1.12M params

Tensor type

BF16

Model tree for March07/Qwen2.5-Coder-32B-terminal-agent-sft-allpass2040

Base model

Qwen/Qwen2.5-32B

Finetuned

Qwen/Qwen2.5-Coder-32B

Finetuned

Qwen/Qwen2.5-Coder-32B-Instruct

Finetuned

(148)

this model