Text Generation
Transformers
Safetensors
Russian
English
ppai
pacperlar
mixture-of-experts
russian
code
small-language-model
custom_code
Instructions to use AdminHugging/PP-Hydrogen with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use AdminHugging/PP-Hydrogen with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="AdminHugging/PP-Hydrogen", trust_remote_code=True)# pip install -U transformers accelerate # Load model directly from transformers import AutoModelForCausalLM model = AutoModelForCausalLM.from_pretrained("AdminHugging/PP-Hydrogen", trust_remote_code=True, device_map="auto") - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- vLLM
How to use AdminHugging/PP-Hydrogen with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "AdminHugging/PP-Hydrogen" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "AdminHugging/PP-Hydrogen", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }'Use Docker
docker model run hf.co/AdminHugging/PP-Hydrogen
- SGLang
How to use AdminHugging/PP-Hydrogen 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 "AdminHugging/PP-Hydrogen" \ --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": "AdminHugging/PP-Hydrogen", "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 "AdminHugging/PP-Hydrogen" \ --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": "AdminHugging/PP-Hydrogen", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }' - Docker Model Runner
How to use AdminHugging/PP-Hydrogen with Docker Model Runner:
docker model run hf.co/AdminHugging/PP-Hydrogen
Upload logs/config_used.yaml with huggingface_hub
Browse files- logs/config_used.yaml +61 -0
logs/config_used.yaml
ADDED
|
@@ -0,0 +1,61 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
root: /kaggle/working/ppai
|
| 2 |
+
seed: 1337
|
| 3 |
+
model:
|
| 4 |
+
base_id: AuroraAI-Research/Aurora2-400M-A20M
|
| 5 |
+
subfolder: bf16
|
| 6 |
+
base_dir: /kaggle/working/ppai/model/base
|
| 7 |
+
tokenizer:
|
| 8 |
+
extend: false
|
| 9 |
+
new_tokens: 16000
|
| 10 |
+
word_share: 0.6
|
| 11 |
+
sample_chars: 20000000
|
| 12 |
+
min_word_freq: 40
|
| 13 |
+
min_word_len: 3
|
| 14 |
+
corpus: ru_wiki
|
| 15 |
+
corpus_file: null
|
| 16 |
+
data:
|
| 17 |
+
total_tokens: 500000000
|
| 18 |
+
seq_len: 1024
|
| 19 |
+
shard_tokens: 50000000
|
| 20 |
+
val_tokens_per_source: 200000
|
| 21 |
+
seed: 1337
|
| 22 |
+
dedup_prefix_chars: 200
|
| 23 |
+
max_scan_per_source: 400000
|
| 24 |
+
use_weights: true
|
| 25 |
+
sources:
|
| 26 |
+
ru_wiki: 0.12
|
| 27 |
+
ru_web: 0.22
|
| 28 |
+
ru_c4: 0.06
|
| 29 |
+
math_numina: 0.12
|
| 30 |
+
math_open: 0.08
|
| 31 |
+
math_meta: 0.05
|
| 32 |
+
code_python: 0.34
|
| 33 |
+
code_csharp: 0.03
|
| 34 |
+
code_rust: 0.03
|
| 35 |
+
en_web: 0.2
|
| 36 |
+
optim:
|
| 37 |
+
lr: 0.0006
|
| 38 |
+
min_lr_ratio: 0.05
|
| 39 |
+
warmup_steps: 500
|
| 40 |
+
weight_decay: 0.1
|
| 41 |
+
beta1: 0.9
|
| 42 |
+
beta2: 0.95
|
| 43 |
+
grad_clip: 1.0
|
| 44 |
+
micro_batch: 12
|
| 45 |
+
grad_accum: 3
|
| 46 |
+
dtype: auto
|
| 47 |
+
params_dtype: fp32
|
| 48 |
+
train:
|
| 49 |
+
train_mode: full
|
| 50 |
+
freeze_embeddings: false
|
| 51 |
+
max_steps: 6780
|
| 52 |
+
save_every: 250
|
| 53 |
+
eval_every: 500
|
| 54 |
+
log_every: 25
|
| 55 |
+
max_hours: 10.5
|
| 56 |
+
resume: true
|
| 57 |
+
hub_repo: AdminHugging/PP-Hydrogen
|
| 58 |
+
hub_private: true
|
| 59 |
+
push_every: 1000
|
| 60 |
+
keep_step_ckpts: 1
|
| 61 |
+
target_tokens: 500000000
|