Instructions to use CausalNLP/gpt2-multilingual-20-arabic-repair_3epochs_lr1e-4_bf16 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use CausalNLP/gpt2-multilingual-20-arabic-repair_3epochs_lr1e-4_bf16 with Transformers:

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

pipe = pipeline("text-generation", model="CausalNLP/gpt2-multilingual-20-arabic-repair_3epochs_lr1e-4_bf16")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("CausalNLP/gpt2-multilingual-20-arabic-repair_3epochs_lr1e-4_bf16")
model = AutoModelForCausalLM.from_pretrained("CausalNLP/gpt2-multilingual-20-arabic-repair_3epochs_lr1e-4_bf16")

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use CausalNLP/gpt2-multilingual-20-arabic-repair_3epochs_lr1e-4_bf16 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "CausalNLP/gpt2-multilingual-20-arabic-repair_3epochs_lr1e-4_bf16"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "CausalNLP/gpt2-multilingual-20-arabic-repair_3epochs_lr1e-4_bf16",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/CausalNLP/gpt2-multilingual-20-arabic-repair_3epochs_lr1e-4_bf16

SGLang

How to use CausalNLP/gpt2-multilingual-20-arabic-repair_3epochs_lr1e-4_bf16 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 "CausalNLP/gpt2-multilingual-20-arabic-repair_3epochs_lr1e-4_bf16" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "CausalNLP/gpt2-multilingual-20-arabic-repair_3epochs_lr1e-4_bf16",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

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 "CausalNLP/gpt2-multilingual-20-arabic-repair_3epochs_lr1e-4_bf16" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "CausalNLP/gpt2-multilingual-20-arabic-repair_3epochs_lr1e-4_bf16",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use CausalNLP/gpt2-multilingual-20-arabic-repair_3epochs_lr1e-4_bf16 with Docker Model Runner:
```
docker model run hf.co/CausalNLP/gpt2-multilingual-20-arabic-repair_3epochs_lr1e-4_bf16
```

gpt2-multilingual-20-arabic-repair_3epochs_lr1e-4_bf16

This model is a fine-tuned version of CausalNLP/gpt2-hf_multilingual-20 on an unknown dataset. It achieves the following results on the evaluation set:

Loss: 3.2206

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 0.0001
train_batch_size: 32
eval_batch_size: 8
seed: 42
gradient_accumulation_steps: 4
total_train_batch_size: 128
optimizer: Use adamw_torch_fused with betas=(0.9,0.95) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: cosine
lr_scheduler_warmup_ratio: 0.1
num_epochs: 3

Training results

Training Loss	Epoch	Step	Validation Loss
3.295	0.0799	500	3.3488
3.2752	0.1597	1000	3.3363
3.2844	0.2396	1500	3.3416
3.3059	0.3194	2000	3.3491
3.335	0.3993	2500	3.3387
3.2845	0.4791	3000	3.3301
3.2458	0.5590	3500	3.3216
3.2689	0.6388	4000	3.3129
3.2737	0.7187	4500	3.3069
3.2604	0.7985	5000	3.3000
3.2255	0.8784	5500	3.2936
3.239	0.9582	6000	3.2882
3.2594	1.0380	6500	3.2832
3.2123	1.1179	7000	3.2780
3.2347	1.1977	7500	3.2726
3.2197	1.2776	8000	3.2671
3.2028	1.3574	8500	3.2615
3.195	1.4373	9000	3.2557
3.2049	1.5171	9500	3.2510
3.1649	1.5970	10000	3.2458
3.1984	1.6768	10500	3.2409
3.2787	1.7567	11000	3.2374
3.1966	1.8365	11500	3.2334
3.166	1.9164	12000	3.2305
3.2029	1.9962	12500	3.2272
3.1629	2.0760	13000	3.2268
3.1115	2.1559	13500	3.2253
3.184	2.2357	14000	3.2239
3.1391	2.3156	14500	3.2229
3.1341	2.3954	15000	3.2220
3.1623	2.4753	15500	3.2214
3.1565	2.5551	16000	3.2211
3.1129	2.6350	16500	3.2208
3.2043	2.7148	17000	3.2207
3.1044	2.7947	17500	3.2206
3.1922	2.8746	18000	3.2206
3.1229	2.9544	18500	3.2206

Framework versions

Transformers 4.57.4
Pytorch 2.9.1+cu128
Datasets 4.4.2
Tokenizers 0.22.2

Downloads last month: 51

Safetensors

Model size

0.2B params

Tensor type

BF16

Model tree for CausalNLP/gpt2-multilingual-20-arabic-repair_3epochs_lr1e-4_bf16

Base model

CausalNLP/gpt2-hf_multilingual-20

Finetuned

(6)

this model

Collection including CausalNLP/gpt2-multilingual-20-arabic-repair_3epochs_lr1e-4_bf16

Multilingual GPT2 Models

Collection

8 items • Updated Mar 2 • 1