Instructions to use Deci/DeciLM-7B-instruct with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Deci/DeciLM-7B-instruct with Transformers:

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

pipe = pipeline("text-generation", model="Deci/DeciLM-7B-instruct", trust_remote_code=True)
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoModelForCausalLM
model = AutoModelForCausalLM.from_pretrained("Deci/DeciLM-7B-instruct", trust_remote_code=True, dtype="auto")

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use Deci/DeciLM-7B-instruct with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Deci/DeciLM-7B-instruct"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Deci/DeciLM-7B-instruct",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Deci/DeciLM-7B-instruct

SGLang

How to use Deci/DeciLM-7B-instruct 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 "Deci/DeciLM-7B-instruct" \
    --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": "Deci/DeciLM-7B-instruct",
		"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 "Deci/DeciLM-7B-instruct" \
        --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": "Deci/DeciLM-7B-instruct",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Deci/DeciLM-7B-instruct with Docker Model Runner:
```
docker model run hf.co/Deci/DeciLM-7B-instruct
```

tomer-deci commited on Dec 17, 2023

Commit

49422ae

1 Parent(s): 64cf789

added support for text-generation pipeline

Browse files

Files changed (3) hide show

config.json +2 -1
modeling_decilm.py +9 -10
version_check.py +3 -3

config.json CHANGED Viewed

@@ -25,5 +25,6 @@
   "use_bfloat16": true,
   "transformers_version": "4.35.2",
   "use_cache": true,
-  "vocab_size": 32000
 }

   "use_bfloat16": true,
   "transformers_version": "4.35.2",
   "use_cache": true,
+  "vocab_size": 32000,
+  "tokenizer_class": "LlamaTokenizer"
 }

modeling_decilm.py CHANGED Viewed

@@ -1,5 +1,5 @@
 # coding=utf-8
-# Copyright and license here
 """ PyTorch DeciLM model."""
 from .version_check import check_transformers_version
@@ -11,18 +11,18 @@ import torch
 import torch.nn.functional as F
 import torch.utils.checkpoint
 from torch import nn
 from .transformers_v4_35_2__modeling_llama import LlamaMLP, LlamaRMSNorm, LlamaAttention, apply_rotary_pos_emb, \
     repeat_kv, LlamaPreTrainedModel, LLAMA_START_DOCSTRING, LlamaDecoderLayer, LlamaForCausalLM, LlamaModel, \
     BaseModelOutputWithPast, LLAMA_INPUTS_DOCSTRING
-from .transformers_v4_35_2__modeling_attn_mask_utils import _prepare_4d_causal_attention_mask
-from transformers.utils import add_start_docstrings, add_start_docstrings_to_model_forward, logging
-from .configuration_decilm import DeciLMConfig
-logger = logging.get_logger(__name__)
 _CONFIG_FOR_DOC = "DeciLMConfig"
 class DeciLMAttention(LlamaAttention):
@@ -61,7 +61,7 @@ class DeciLMAttention(LlamaAttention):
             past_key_value: Optional[Tuple[torch.Tensor]] = None,
             output_attentions: bool = False,
             use_cache: bool = False,
-            # padding_mask: Optional[torch.LongTensor] = None,
     ) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
         bsz, q_len, _ = hidden_states.size()
         is_decode = past_key_value is not None
@@ -135,8 +135,7 @@ class DeciLMAttention(LlamaAttention):
         else:
             attn_output = self.o_proj(attn_output)
-        if not output_attentions:
-            attn_weights = None
         return attn_output, attn_weights, past_key_value

 # coding=utf-8
+# Copyright and license in the repo.
 """ PyTorch DeciLM model."""
 from .version_check import check_transformers_version
 import torch.nn.functional as F
 import torch.utils.checkpoint
 from torch import nn
+from transformers.models.auto.modeling_auto import MODEL_FOR_CAUSAL_LM_MAPPING_NAMES
+from transformers.utils import add_start_docstrings, add_start_docstrings_to_model_forward, logging
+from .configuration_decilm import DeciLMConfig
+from .transformers_v4_35_2__modeling_attn_mask_utils import _prepare_4d_causal_attention_mask
 from .transformers_v4_35_2__modeling_llama import LlamaMLP, LlamaRMSNorm, LlamaAttention, apply_rotary_pos_emb, \
     repeat_kv, LlamaPreTrainedModel, LLAMA_START_DOCSTRING, LlamaDecoderLayer, LlamaForCausalLM, LlamaModel, \
     BaseModelOutputWithPast, LLAMA_INPUTS_DOCSTRING
+MODEL_FOR_CAUSAL_LM_MAPPING_NAMES["deci"] = "DeciLMForCausalLM"
 _CONFIG_FOR_DOC = "DeciLMConfig"
+logger = logging.get_logger(__name__)
 class DeciLMAttention(LlamaAttention):
             past_key_value: Optional[Tuple[torch.Tensor]] = None,
             output_attentions: bool = False,
             use_cache: bool = False,
+            **kwargs,
     ) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
         bsz, q_len, _ = hidden_states.size()
         is_decode = past_key_value is not None
         else:
             attn_output = self.o_proj(attn_output)
+        attn_weights = None
         return attn_output, attn_weights, past_key_value

version_check.py CHANGED Viewed

@@ -1,11 +1,11 @@
 import transformers
 from packaging import version
-VERSION = "4.35.2"
 def check_transformers_version():
-    if version.parse(transformers.__version__) < version.parse(VERSION):
         raise ImportError(
-            f"You are using transformers=={transformers.__version__}, but transformers>={VERSION} is required to use DeciLM. Please upgrade transformers."
         )

 import transformers
 from packaging import version
+MIN_VERSION = "4.35.2"
 def check_transformers_version():
+    if version.parse(transformers.__version__) < version.parse(MIN_VERSION):
         raise ImportError(
+            f"You are using transformers=={transformers.__version__}, but transformers>={MIN_VERSION} is required to use DeciLM. Please upgrade transformers."
         )