Instructions to use RichardLu/Mistral7b_AE_res with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use RichardLu/Mistral7b_AE_res with Transformers:
# Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("RichardLu/Mistral7b_AE_res", device_map="auto") - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- Unsloth Studio
How to use RichardLu/Mistral7b_AE_res with Unsloth Studio:
Install Unsloth Studio (macOS, Linux, WSL)
curl -fsSL https://unsloth.ai/install.sh | sh # Run unsloth studio unsloth studio -H 0.0.0.0 -p 8888 # Then open http://localhost:8888 in your browser # Search for RichardLu/Mistral7b_AE_res to start chatting
Install Unsloth Studio (Windows)
irm https://unsloth.ai/install.ps1 | iex # Run unsloth studio unsloth studio -H 0.0.0.0 -p 8888 # Then open http://localhost:8888 in your browser # Search for RichardLu/Mistral7b_AE_res to start chatting
Using HuggingFace Spaces for Unsloth
# No setup required # Open https://huggingface.co/spaces/unsloth/studio in your browser # Search for RichardLu/Mistral7b_AE_res to start chatting
Load model with FastModel
pip install unsloth from unsloth import FastModel model, tokenizer = FastModel.from_pretrained( model_name="RichardLu/Mistral7b_AE_res", max_seq_length=2048, )
Update handler.py
Browse files- handler.py +10 -6
handler.py
CHANGED
|
@@ -9,20 +9,23 @@ class EndpointHandler:
|
|
| 9 |
# Retrieve the Hugging Face token from the environment variable.
|
| 10 |
hf_token = os.getenv("HF_TOKEN", None)
|
| 11 |
|
| 12 |
-
#
|
|
|
|
|
|
|
|
|
|
|
|
|
| 13 |
max_seq_length = 2048
|
| 14 |
dtype = None # Auto-detect data type; adjust if needed.
|
| 15 |
load_in_4bit = True # Use 4-bit quantization to reduce memory usage.
|
| 16 |
|
| 17 |
# Load the model and tokenizer.
|
| 18 |
-
# If a model_dir is provided by the endpoint, it will load from that path;
|
| 19 |
-
# otherwise, it uses a default UnsLoth Mistral model.
|
| 20 |
self.model, self.tokenizer = FastLanguageModel.from_pretrained(
|
| 21 |
-
model_name=
|
| 22 |
max_seq_length=max_seq_length,
|
| 23 |
dtype=dtype,
|
| 24 |
load_in_4bit=load_in_4bit,
|
| 25 |
-
token=hf_token
|
|
|
|
| 26 |
)
|
| 27 |
|
| 28 |
# Switch the model to inference mode.
|
|
@@ -30,7 +33,8 @@ class EndpointHandler:
|
|
| 30 |
|
| 31 |
# Define the instruction text with examples for aspect extraction.
|
| 32 |
self.instructabsa_instruction = (
|
| 33 |
-
"Definition: The output will be the aspects (both implicit and explicit) which have an associated opinion that are extracted from the input text.
|
|
|
|
| 34 |
"Positive example 1-\n"
|
| 35 |
"input: With the great variety on the menu, I eat here often and never get bored.\n"
|
| 36 |
"output: menu\n"
|
|
|
|
| 9 |
# Retrieve the Hugging Face token from the environment variable.
|
| 10 |
hf_token = os.getenv("HF_TOKEN", None)
|
| 11 |
|
| 12 |
+
# If the endpoint provides '/repository', override with the intended model ID.
|
| 13 |
+
if model_dir == "/repository" or model_dir.strip() == "":
|
| 14 |
+
model_dir = "RichardLu/Mistral7b_AE_res"
|
| 15 |
+
|
| 16 |
+
# Configuration for inference.
|
| 17 |
max_seq_length = 2048
|
| 18 |
dtype = None # Auto-detect data type; adjust if needed.
|
| 19 |
load_in_4bit = True # Use 4-bit quantization to reduce memory usage.
|
| 20 |
|
| 21 |
# Load the model and tokenizer.
|
|
|
|
|
|
|
| 22 |
self.model, self.tokenizer = FastLanguageModel.from_pretrained(
|
| 23 |
+
model_name=model_dir,
|
| 24 |
max_seq_length=max_seq_length,
|
| 25 |
dtype=dtype,
|
| 26 |
load_in_4bit=load_in_4bit,
|
| 27 |
+
token=hf_token, # Use the HF token if provided.
|
| 28 |
+
trust_remote_code=True # Enable if your model requires remote code.
|
| 29 |
)
|
| 30 |
|
| 31 |
# Switch the model to inference mode.
|
|
|
|
| 33 |
|
| 34 |
# Define the instruction text with examples for aspect extraction.
|
| 35 |
self.instructabsa_instruction = (
|
| 36 |
+
"Definition: The output will be the aspects (both implicit and explicit) which have an associated opinion that are extracted from the input text. "
|
| 37 |
+
"In cases where there are no aspects the output should be noaspectterm.\n"
|
| 38 |
"Positive example 1-\n"
|
| 39 |
"input: With the great variety on the menu, I eat here often and never get bored.\n"
|
| 40 |
"output: menu\n"
|