Instructions to use andreaskoepf/pythia-12b-pre-2000 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use andreaskoepf/pythia-12b-pre-2000 with Transformers:

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

pipe = pipeline("text-generation", model="andreaskoepf/pythia-12b-pre-2000")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("andreaskoepf/pythia-12b-pre-2000")
model = AutoModelForCausalLM.from_pretrained("andreaskoepf/pythia-12b-pre-2000")

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use andreaskoepf/pythia-12b-pre-2000 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "andreaskoepf/pythia-12b-pre-2000"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "andreaskoepf/pythia-12b-pre-2000",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/andreaskoepf/pythia-12b-pre-2000

SGLang

How to use andreaskoepf/pythia-12b-pre-2000 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 "andreaskoepf/pythia-12b-pre-2000" \
    --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": "andreaskoepf/pythia-12b-pre-2000",
		"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 "andreaskoepf/pythia-12b-pre-2000" \
        --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": "andreaskoepf/pythia-12b-pre-2000",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use andreaskoepf/pythia-12b-pre-2000 with Docker Model Runner:
```
docker model run hf.co/andreaskoepf/pythia-12b-pre-2000
```

New run with DeepSpeed 0.7.7 and transformers 4.26.1, (using WarmupDecayLR, probably not optimal).

wandb run: https://wandb.ai/open-assistant/supervised-finetuning/runs/bxyaxo4v Checkpoint: 2000 steps ~48% 1st epoch

Datasets:

pretrain:
  use_custom_sampler: true
  sort_by_length: false
  datasets:
    - joke
    - webgpt:
        val_split: 0.1
    - gpt4all:
        val_split: 0.01
    - alpaca:
        val_split: 0.025
    - code_alpaca:
        val_split: 0.05
    - minimath
    - humaneval_mbpp_codegen_qa
    - humaneval_mbpp_testgen_qa
    - grade_school_math_instructions
    - recipes
    - cmu_wiki_qa
    #- youtube_subs_howto100m # uses incompatible column names
    #- ubuntu_dialogue_qa # fails to load
    - oa_wiki_qa_bart_10000row
    - prosocial_dialogue:
        fraction: 0.1
    - explain_prosocial:
        fraction: 0.05

Pythia:

pythia-12b:
  dtype: fp16
  log_dir: "pythia_log_12b"
  learning_rate: 6e-6
  model_name: EleutherAI/pythia-12b-deduped
  output_dir: pythia_model_12b
  weight_decay: 0.0
  max_length: 2048
  use_flash_attention: true
  deepspeed_config: configs/zero_conf2.json
  warmup_steps: 50
  gradient_checkpointing: true
  gradient_accumulation_steps: 2
  per_device_train_batch_size: 8 
  per_device_eval_batch_size: 5
  eval_steps: 200
  save_steps: 500
  num_train_epochs: 2 
  save_total_limit: 2

zero_conf2.json:

{
    "fp16": {
        "enabled": "auto",
        "loss_scale": 0,
        "loss_scale_window": 1000,
        "initial_scale_power": 16,
        "hysteresis": 2,
        "min_loss_scale": 1
    },
    "bf16": {
        "enabled": "auto"
    },
    "optimizer": {
        "type": "AdamW",
        "params": {
            "lr": "auto",
            "betas": "auto",
            "eps": "auto",
            "weight_decay": "auto"
        }
    },
    "scheduler": {
        "type": "WarmupDecayLR",
        "params": {
            "warmup_min_lr": "auto",
            "warmup_max_lr": "auto",
            "warmup_num_steps": "auto",
            "total_num_steps": "auto"
        }
    },
    "zero_optimization": {
        "stage": 2,
        "allgather_partitions": true,
        "allgather_bucket_size": 1000000000.0,
        "overlap_comm": false,
        "reduce_scatter": true,
        "reduce_bucket_size": 1000000000.0,
        "contiguous_gradients": true
    },
    "gradient_accumulation_steps": "auto",
    "gradient_clipping": "auto",
    "steps_per_print": 2000,
    "train_batch_size": "auto",
    "train_micro_batch_size_per_gpu": "auto",
    "wall_clock_breakdown": false
}

Downloads last month: 11