Instructions to use Devnexai/gemma4-code-assistant with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Devnexai/gemma4-code-assistant with PEFT:

from peft import PeftModel
from transformers import AutoModelForCausalLM

base_model = AutoModelForCausalLM.from_pretrained("google/gemma-4-E2B")
model = PeftModel.from_pretrained(base_model, "Devnexai/gemma4-code-assistant")

Transformers

How to use Devnexai/gemma4-code-assistant with Transformers:

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

pipe = pipeline("text-generation", model="Devnexai/gemma4-code-assistant")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("Devnexai/gemma4-code-assistant", dtype="auto")

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use Devnexai/gemma4-code-assistant with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Devnexai/gemma4-code-assistant"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Devnexai/gemma4-code-assistant",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Devnexai/gemma4-code-assistant

SGLang

How to use Devnexai/gemma4-code-assistant 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 "Devnexai/gemma4-code-assistant" \
    --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": "Devnexai/gemma4-code-assistant",
		"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 "Devnexai/gemma4-code-assistant" \
        --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": "Devnexai/gemma4-code-assistant",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Devnexai/gemma4-code-assistant with Docker Model Runner:
```
docker model run hf.co/Devnexai/gemma4-code-assistant
```

gemma4-code-assistant / checkpoint-1875 /adapter_config.json

nzros

Upload folder using huggingface_hub

2ee8954 verified about 1 month ago

raw

history blame contribute delete

9.82 kB

	{
	"alora_invocation_tokens": null,
	"alpha_pattern": {},
	"arrow_config": null,
	"auto_mapping": null,
	"base_model_name_or_path": "google/gemma-4-E2B",
	"bias": "none",
	"corda_config": null,
	"ensure_weight_tying": false,
	"eva_config": null,
	"exclude_modules": null,
	"fan_in_fan_out": false,
	"inference_mode": true,
	"init_lora_weights": true,
	"layer_replication": null,
	"layers_pattern": null,
	"layers_to_transform": null,
	"loftq_config": {},
	"lora_alpha": 32,
	"lora_bias": false,
	"lora_dropout": 0.05,
	"megatron_config": null,
	"megatron_core": "megatron.core",
	"modules_to_save": [
	"lm_head",
	"embed_tokens"
	],
	"peft_type": "LORA",
	"peft_version": "0.18.1",
	"qalora_group_size": 16,
	"r": 16,
	"rank_pattern": {},
	"revision": null,
	"target_modules": [
	"26.self_attn.q_proj",
	"language_model.layers.11.mlp.gate_proj",
	"34.mlp.gate_proj",
	"23.mlp.gate_proj",
	"language_model.layers.5.mlp.gate_proj",
	"input_proj",
	"language_model.layers.12.self_attn.q_proj",
	"32.mlp.down_proj",
	"20.mlp.up_proj",
	"language_model.layers.4.self_attn.v_proj",
	"20.self_attn.k_proj",
	"input_proj_linear",
	"24.mlp.gate_proj",
	"29.self_attn.k_proj",
	"21.self_attn.v_proj",
	"21.mlp.down_proj",
	"16.self_attn.v_proj",
	"language_model.layers.13.mlp.up_proj",
	"24.mlp.down_proj",
	"22.mlp.up_proj",
	"language_model.layers.9.self_attn.k_proj",
	"26.mlp.up_proj",
	"language_model.layers.7.mlp.down_proj",
	"language_model.layers.2.mlp.gate_proj",
	"language_model.layers.13.self_attn.k_proj",
	"25.self_attn.q_proj",
	"per_layer_model_projection",
	"18.mlp.up_proj",
	"language_model.layers.13.self_attn.v_proj",
	"34.mlp.down_proj",
	"23.mlp.up_proj",
	"23.self_attn.o_proj",
	"24.self_attn.v_proj",
	"language_model.layers.6.self_attn.q_proj",
	"29.self_attn.q_proj",
	"19.self_attn.k_proj",
	"24.self_attn.k_proj",
	"language_model.layers.5.self_attn.q_proj",
	"32.mlp.gate_proj",
	"language_model.layers.4.mlp.gate_proj",
	"33.mlp.gate_proj",
	"19.self_attn.o_proj",
	"34.self_attn.v_proj",
	"25.mlp.down_proj",
	"language_model.layers.8.mlp.up_proj",
	"language_model.layers.12.mlp.down_proj",
	"language_model.layers.3.mlp.down_proj",
	"language_model.layers.7.self_attn.k_proj",
	"language_model.layers.2.mlp.up_proj",
	"language_model.layers.13.mlp.down_proj",
	"28.mlp.gate_proj",
	"language_model.layers.15.self_attn.v_proj",
	"20.self_attn.q_proj",
	"per_layer_projection",
	"27.mlp.gate_proj",
	"language_model.layers.11.self_attn.o_proj",
	"language_model.layers.15.mlp.gate_proj",
	"30.self_attn.o_proj",
	"32.self_attn.o_proj",
	"27.mlp.down_proj",
	"language_model.layers.1.mlp.gate_proj",
	"21.self_attn.o_proj",
	"32.self_attn.q_proj",
	"language_model.layers.7.self_attn.v_proj",
	"21.mlp.gate_proj",
	"16.self_attn.k_proj",
	"20.mlp.down_proj",
	"16.mlp.gate_proj",
	"20.mlp.gate_proj",
	"language_model.layers.7.mlp.gate_proj",
	"34.self_attn.q_proj",
	"31.mlp.down_proj",
	"language_model.layers.8.mlp.gate_proj",
	"language_model.layers.9.mlp.up_proj",
	"language_model.layers.15.self_attn.o_proj",
	"28.self_attn.o_proj",
	"32.self_attn.v_proj",
	"32.self_attn.k_proj",
	"language_model.layers.4.self_attn.q_proj",
	"28.self_attn.v_proj",
	"28.self_attn.k_proj",
	"26.self_attn.k_proj",
	"34.self_attn.k_proj",
	"16.mlp.up_proj",
	"21.self_attn.q_proj",
	"17.self_attn.q_proj",
	"language_model.layers.11.self_attn.k_proj",
	"24.mlp.up_proj",
	"23.self_attn.q_proj",
	"language_model.layers.3.mlp.up_proj",
	"language_model.layers.12.self_attn.k_proj",
	"language_model.layers.1.mlp.down_proj",
	"20.self_attn.v_proj",
	"language_model.layers.11.mlp.up_proj",
	"language_model.layers.0.mlp.gate_proj",
	"language_model.layers.5.self_attn.v_proj",
	"language_model.layers.6.self_attn.o_proj",
	"language_model.layers.11.mlp.down_proj",
	"31.mlp.gate_proj",
	"31.self_attn.q_proj",
	"22.self_attn.v_proj",
	"31.self_attn.v_proj",
	"25.mlp.up_proj",
	"language_model.layers.9.self_attn.q_proj",
	"27.self_attn.o_proj",
	"17.mlp.up_proj",
	"25.self_attn.k_proj",
	"language_model.layers.1.self_attn.o_proj",
	"19.mlp.down_proj",
	"language_model.layers.14.self_attn.q_proj",
	"31.self_attn.k_proj",
	"language_model.layers.14.mlp.down_proj",
	"language_model.layers.2.self_attn.v_proj",
	"language_model.layers.9.mlp.down_proj",
	"26.mlp.down_proj",
	"30.self_attn.q_proj",
	"31.self_attn.o_proj",
	"18.self_attn.v_proj",
	"33.self_attn.q_proj",
	"language_model.layers.2.self_attn.q_proj",
	"30.mlp.up_proj",
	"language_model.layers.4.mlp.down_proj",
	"29.mlp.down_proj",
	"19.self_attn.q_proj",
	"language_model.layers.1.self_attn.v_proj",
	"language_model.layers.10.self_attn.o_proj",
	"27.self_attn.k_proj",
	"embedding_projection",
	"24.self_attn.q_proj",
	"33.self_attn.v_proj",
	"29.mlp.up_proj",
	"25.self_attn.v_proj",
	"language_model.layers.11.self_attn.q_proj",
	"33.self_attn.k_proj",
	"language_model.layers.5.self_attn.o_proj",
	"language_model.layers.4.self_attn.k_proj",
	"language_model.layers.3.self_attn.o_proj",
	"21.self_attn.k_proj",
	"language_model.layers.8.mlp.down_proj",
	"language_model.layers.2.self_attn.o_proj",
	"20.self_attn.o_proj",
	"language_model.layers.0.self_attn.k_proj",
	"linear",
	"language_model.layers.10.self_attn.v_proj",
	"17.mlp.down_proj",
	"21.mlp.up_proj",
	"language_model.layers.2.mlp.down_proj",
	"22.mlp.gate_proj",
	"language_model.layers.9.self_attn.o_proj",
	"language_model.layers.0.self_attn.q_proj",
	"language_model.layers.14.self_attn.k_proj",
	"18.mlp.gate_proj",
	"language_model.layers.3.self_attn.q_proj",
	"language_model.layers.7.self_attn.q_proj",
	"language_model.layers.15.mlp.up_proj",
	"34.self_attn.o_proj",
	"language_model.layers.8.self_attn.q_proj",
	"language_model.layers.10.mlp.up_proj",
	"language_model.layers.14.mlp.up_proj",
	"29.mlp.gate_proj",
	"language_model.layers.1.self_attn.k_proj",
	"language_model.layers.13.self_attn.o_proj",
	"25.self_attn.o_proj",
	"17.self_attn.v_proj",
	"30.mlp.gate_proj",
	"26.self_attn.v_proj",
	"language_model.layers.3.self_attn.v_proj",
	"19.self_attn.v_proj",
	"language_model.layers.0.mlp.down_proj",
	"language_model.layers.3.self_attn.k_proj",
	"18.self_attn.o_proj",
	"29.self_attn.o_proj",
	"25.mlp.gate_proj",
	"language_model.layers.9.mlp.gate_proj",
	"16.self_attn.o_proj",
	"language_model.layers.12.self_attn.o_proj",
	"33.self_attn.o_proj",
	"output_proj",
	"22.self_attn.o_proj",
	"17.self_attn.o_proj",
	"23.self_attn.v_proj",
	"language_model.layers.12.mlp.gate_proj",
	"22.self_attn.q_proj",
	"relative_k_proj",
	"30.mlp.down_proj",
	"22.mlp.down_proj",
	"31.mlp.up_proj",
	"18.self_attn.k_proj",
	"language_model.layers.1.mlp.up_proj",
	"28.mlp.down_proj",
	"language_model.layers.15.self_attn.q_proj",
	"language_model.layers.10.mlp.gate_proj",
	"language_model.layers.0.self_attn.v_proj",
	"23.mlp.down_proj",
	"29.self_attn.v_proj",
	"language_model.layers.8.self_attn.v_proj",
	"22.self_attn.k_proj",
	"17.mlp.gate_proj",
	"24.self_attn.o_proj",
	"16.mlp.down_proj",
	"33.mlp.down_proj",
	"language_model.layers.2.self_attn.k_proj",
	"23.self_attn.k_proj",
	"30.self_attn.v_proj",
	"language_model.layers.4.self_attn.o_proj",
	"language_model.layers.1.self_attn.q_proj",
	"language_model.layers.7.mlp.up_proj",
	"language_model.layers.5.self_attn.k_proj",
	"19.mlp.up_proj",
	"per_layer_input_gate",
	"language_model.layers.0.mlp.up_proj",
	"30.self_attn.k_proj",
	"language_model.layers.8.self_attn.k_proj",
	"18.self_attn.q_proj",
	"language_model.layers.4.mlp.up_proj",
	"19.mlp.gate_proj",
	"language_model.layers.0.self_attn.o_proj",
	"language_model.layers.12.self_attn.v_proj",
	"34.mlp.up_proj",
	"language_model.layers.13.self_attn.q_proj",
	"27.mlp.up_proj",
	"26.mlp.gate_proj",
	"language_model.layers.6.self_attn.k_proj",
	"26.self_attn.o_proj",
	"language_model.layers.10.self_attn.q_proj",
	"language_model.layers.10.self_attn.k_proj",
	"language_model.layers.10.mlp.down_proj",
	"33.mlp.up_proj",
	"language_model.layers.6.self_attn.v_proj",
	"16.self_attn.q_proj",
	"27.self_attn.v_proj",
	"28.mlp.up_proj",
	"language_model.layers.15.mlp.down_proj",
	"language_model.layers.14.self_attn.o_proj",
	"language_model.layers.5.mlp.up_proj",
	"language_model.layers.13.mlp.gate_proj",
	"language_model.layers.14.mlp.gate_proj",
	"language_model.layers.11.self_attn.v_proj",
	"language_model.layers.14.self_attn.v_proj",
	"18.mlp.down_proj",
	"17.self_attn.k_proj",
	"language_model.layers.7.self_attn.o_proj",
	"32.mlp.up_proj",
	"language_model.layers.6.mlp.up_proj",
	"language_model.layers.9.self_attn.v_proj",
	"language_model.layers.3.mlp.gate_proj",
	"language_model.layers.6.mlp.down_proj",
	"language_model.layers.12.mlp.up_proj",
	"language_model.layers.6.mlp.gate_proj",
	"28.self_attn.q_proj",
	"language_model.layers.15.self_attn.k_proj",
	"language_model.layers.8.self_attn.o_proj",
	"language_model.layers.5.mlp.down_proj",
	"27.self_attn.q_proj"
	],
	"target_parameters": null,
	"task_type": "CAUSAL_LM",
	"trainable_token_indices": null,
	"use_dora": false,
	"use_qalora": false,
	"use_rslora": false
	}