Instructions to use lightonai/RITA_l with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use lightonai/RITA_l with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="lightonai/RITA_l", trust_remote_code=True)# pip install -U transformers accelerate # Load model directly from transformers import AutoModelForCausalLM model = AutoModelForCausalLM.from_pretrained("lightonai/RITA_l", trust_remote_code=True, device_map="auto") - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- vLLM
How to use lightonai/RITA_l with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "lightonai/RITA_l" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "lightonai/RITA_l", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }'Use Docker
docker model run hf.co/lightonai/RITA_l
- SGLang
How to use lightonai/RITA_l 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 "lightonai/RITA_l" \ --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": "lightonai/RITA_l", "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 "lightonai/RITA_l" \ --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": "lightonai/RITA_l", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }' - Docker Model Runner
How to use lightonai/RITA_l with Docker Model Runner:
docker model run hf.co/lightonai/RITA_l
Commit ·
1c2046a
1
Parent(s): 5b0200e
add model
Browse files- config.json +7 -11
- pytorch_model.bin +2 -2
- rita_configuration.py +5 -8
- rita_modeling.py +4 -5
config.json
CHANGED
|
@@ -1,26 +1,22 @@
|
|
| 1 |
{
|
| 2 |
-
"_name_or_path": "
|
| 3 |
"architectures": [
|
| 4 |
"RITAModel"
|
| 5 |
],
|
| 6 |
"auto_map": {
|
| 7 |
"AutoConfig": "rita_configuration.RITAConfig",
|
| 8 |
-
"AutoModel": "rita_modeling.RITAModel"
|
|
|
|
| 9 |
},
|
| 10 |
-
"bos_token_id": [
|
| 11 |
-
[
|
| 12 |
-
50256
|
| 13 |
-
]
|
| 14 |
-
],
|
| 15 |
"d_feedforward": 6144,
|
| 16 |
"d_model": 1536,
|
| 17 |
"dropout": 0.0,
|
| 18 |
-
"eos_token_id":
|
| 19 |
"max_seq_len": 1024,
|
| 20 |
-
"model_type": "
|
| 21 |
"num_heads": 24,
|
| 22 |
"num_layers": 24,
|
| 23 |
-
"torch_dtype": "
|
| 24 |
"transformers_version": "4.18.0",
|
| 25 |
-
"vocab_size":
|
| 26 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"_name_or_path": "nz/RITA_l",
|
| 3 |
"architectures": [
|
| 4 |
"RITAModel"
|
| 5 |
],
|
| 6 |
"auto_map": {
|
| 7 |
"AutoConfig": "rita_configuration.RITAConfig",
|
| 8 |
+
"AutoModel": "rita_modeling.RITAModel",
|
| 9 |
+
"AutoModelForCausalLM": "rita_modeling.RITAModel"
|
| 10 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 11 |
"d_feedforward": 6144,
|
| 12 |
"d_model": 1536,
|
| 13 |
"dropout": 0.0,
|
| 14 |
+
"eos_token_id": 2,
|
| 15 |
"max_seq_len": 1024,
|
| 16 |
+
"model_type": "rita",
|
| 17 |
"num_heads": 24,
|
| 18 |
"num_layers": 24,
|
| 19 |
+
"torch_dtype": "float16",
|
| 20 |
"transformers_version": "4.18.0",
|
| 21 |
+
"vocab_size": 26
|
| 22 |
}
|
pytorch_model.bin
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7fb144083615545fb1c5b31be836ba1adfffd8c0e4fff1cef287b9d3108502a1
|
| 3 |
+
size 1360125515
|
rita_configuration.py
CHANGED
|
@@ -1,26 +1,24 @@
|
|
| 1 |
-
|
| 2 |
from transformers.configuration_utils import PretrainedConfig
|
| 3 |
from transformers.utils import logging
|
| 4 |
|
| 5 |
logger = logging.get_logger(__name__)
|
| 6 |
|
| 7 |
class RITAConfig(PretrainedConfig):
|
| 8 |
-
model_type = "
|
| 9 |
|
| 10 |
def __init__(
|
| 11 |
self,
|
| 12 |
-
vocab_size=
|
| 13 |
d_model=768,
|
| 14 |
num_layers=12,
|
| 15 |
max_seq_len=1024,
|
| 16 |
num_heads=12,
|
| 17 |
dropout=0.,
|
| 18 |
ff_ratio=4,
|
| 19 |
-
|
| 20 |
-
eos_token_id=50256, # TODO
|
| 21 |
**kwargs,
|
| 22 |
):
|
| 23 |
-
super().__init__(
|
| 24 |
self.vocab_size = vocab_size
|
| 25 |
self.d_model = d_model
|
| 26 |
self.num_heads = num_heads
|
|
@@ -28,5 +26,4 @@ class RITAConfig(PretrainedConfig):
|
|
| 28 |
self.num_layers = num_layers
|
| 29 |
self.max_seq_len=max_seq_len
|
| 30 |
self.dropout = dropout
|
| 31 |
-
self.
|
| 32 |
-
self.eos_token_id=eos_token_id
|
|
|
|
|
|
|
| 1 |
from transformers.configuration_utils import PretrainedConfig
|
| 2 |
from transformers.utils import logging
|
| 3 |
|
| 4 |
logger = logging.get_logger(__name__)
|
| 5 |
|
| 6 |
class RITAConfig(PretrainedConfig):
|
| 7 |
+
model_type = "rita"
|
| 8 |
|
| 9 |
def __init__(
|
| 10 |
self,
|
| 11 |
+
vocab_size=26,
|
| 12 |
d_model=768,
|
| 13 |
num_layers=12,
|
| 14 |
max_seq_len=1024,
|
| 15 |
num_heads=12,
|
| 16 |
dropout=0.,
|
| 17 |
ff_ratio=4,
|
| 18 |
+
eos_token_id=2,
|
|
|
|
| 19 |
**kwargs,
|
| 20 |
):
|
| 21 |
+
super().__init__(eos_token_id=eos_token_id, **kwargs)
|
| 22 |
self.vocab_size = vocab_size
|
| 23 |
self.d_model = d_model
|
| 24 |
self.num_heads = num_heads
|
|
|
|
| 26 |
self.num_layers = num_layers
|
| 27 |
self.max_seq_len=max_seq_len
|
| 28 |
self.dropout = dropout
|
| 29 |
+
self.eos_token_id=eos_token_id
|
|
|
rita_modeling.py
CHANGED
|
@@ -222,10 +222,10 @@ class RITAModel(PreTrainedModel):
|
|
| 222 |
self.final_norm = nn.LayerNorm(config.d_model)
|
| 223 |
self.projector = nn.Linear(config.d_model, config.vocab_size, bias = False)
|
| 224 |
|
| 225 |
-
def forward(self,
|
| 226 |
-
x = self.embedding(
|
| 227 |
if attn_mask == None:
|
| 228 |
-
attn_mask = (torch.triu(torch.ones(
|
| 229 |
for layer in self.layers:
|
| 230 |
x = layer(x, attn_mask=attn_mask, padding_mask=padding_mask)
|
| 231 |
x = self.final_norm(x) # N x L x D
|
|
@@ -246,5 +246,4 @@ class RITAModel(PreTrainedModel):
|
|
| 246 |
return self.projector
|
| 247 |
|
| 248 |
def set_output_embeddings(self, new_projector):
|
| 249 |
-
|
| 250 |
-
|
|
|
|
| 222 |
self.final_norm = nn.LayerNorm(config.d_model)
|
| 223 |
self.projector = nn.Linear(config.d_model, config.vocab_size, bias = False)
|
| 224 |
|
| 225 |
+
def forward(self, input_ids, attn_mask=None, padding_mask=None, return_hidden=False) -> torch.FloatTensor:
|
| 226 |
+
x = self.embedding(input_ids) # N x L x D
|
| 227 |
if attn_mask == None:
|
| 228 |
+
attn_mask = (torch.triu(torch.ones(input_ids.size(1), input_ids.size(1))) == 0).transpose(0, 1).contiguous().to(input_ids.device)
|
| 229 |
for layer in self.layers:
|
| 230 |
x = layer(x, attn_mask=attn_mask, padding_mask=padding_mask)
|
| 231 |
x = self.final_norm(x) # N x L x D
|
|
|
|
| 246 |
return self.projector
|
| 247 |
|
| 248 |
def set_output_embeddings(self, new_projector):
|
| 249 |
+
self.projector = new_projector
|
|
|