Instructions to use tomg-group-umd/step-00011904-recurrence_full_512_0 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use tomg-group-umd/step-00011904-recurrence_full_512_0 with Transformers:

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

pipe = pipeline("text-generation", model="tomg-group-umd/step-00011904-recurrence_full_512_0", trust_remote_code=True)
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoModelForCausalLM
model = AutoModelForCausalLM.from_pretrained("tomg-group-umd/step-00011904-recurrence_full_512_0", trust_remote_code=True, dtype="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use tomg-group-umd/step-00011904-recurrence_full_512_0 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "tomg-group-umd/step-00011904-recurrence_full_512_0"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "tomg-group-umd/step-00011904-recurrence_full_512_0",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/tomg-group-umd/step-00011904-recurrence_full_512_0

SGLang

How to use tomg-group-umd/step-00011904-recurrence_full_512_0 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 "tomg-group-umd/step-00011904-recurrence_full_512_0" \
    --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": "tomg-group-umd/step-00011904-recurrence_full_512_0",
		"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 "tomg-group-umd/step-00011904-recurrence_full_512_0" \
        --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": "tomg-group-umd/step-00011904-recurrence_full_512_0",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use tomg-group-umd/step-00011904-recurrence_full_512_0 with Docker Model Runner:
```
docker model run hf.co/tomg-group-umd/step-00011904-recurrence_full_512_0
```

step-00011904-recurrence_full_512_0 / config.json

JonasGeiping

Upload RavenForCausalLM

fef90c9 verified over 1 year ago

Raw

History Blame Contribute Delete

1.37 kB

	{
	"activation_checkpoint_impl": "per-iteration",
	"architecture_class_name": "RecurrentGPT",
	"architectures": [
	"RavenForCausalLM"
	],
	"auto_map": {
	"AutoConfig": "raven_config_minimal.RavenConfig",
	"AutoModelForCausalLM": "raven_modeling_minimal.RavenForCausalLM"
	},
	"bias": false,
	"block_class_name": "SandwichBlock",
	"block_size": 4096,
	"effective_expected_depth": 132,
	"head_dim": 96,
	"init_orthogonal": false,
	"init_strategy": "takase",
	"init_values": {
	"embed_scale": 72.6636084983398,
	"embedding": 0.008703882797784892,
	"out_proj": 0.0005356869554443541,
	"std": 0.008703882797784892
	},
	"injection_type": "linear",
	"intermediate_size": 17920,
	"mean_backprop_depth": 8,
	"mean_recurrence": 32,
	"mlp_class_name": "GatedMLP",
	"model_type": "huginn_raven",
	"n_embd": 5280,
	"n_heads": 55,
	"n_layers": 8,
	"n_layers_in_coda": 2,
	"n_layers_in_prelude": 2,
	"n_layers_in_recurrent_block": 4,
	"nonlin_name": "SiLU",
	"norm_class_name": "RMSNorm_llama",
	"norm_eps": 1e-06,
	"num_key_value_heads": 55,
	"padded_vocab_size": 65536,
	"padding_multiple": 4096,
	"qk_bias": true,
	"rope_base": 50000,
	"sampling_scheme": "poisson-lognormal-filling",
	"state_init": "like-init",
	"tie_embeddings": true,
	"torch_dtype": "float32",
	"transformers_version": "4.44.2",
	"vocab_size": 65536
	}