metadata
license: other
license_name: glm-4
license_link: https://huggingface.co/THUDM/glm-4-9b-chat/blob/main/LICENSE
language:
- en
- zh
library_name: transformers
pipeline_tag: text-generation
base_model: THUDM/glm-4-9b-chat
tags:
- Mental Health
- Chatbot
- LLM
- llama-factory
- EMOLLM
Update
The model is now following the update from GLM-4-9B-Chat and now requires transformers>=4.44.0. Please update your dependencies accordingly.
Also follow the dependencies it before using
Introduction
This model is GLM-4-9B-Chat, fine-tuned with various datasets to focus on mental health care.
Since it is fine-tuned with a Chinese dataset, please use it in Chinese, even though the base model supports English text.
Dataset
- Smile dataset
- SoulChat
- single_turn_dataset_1 from EMOLLM
- [self-defined role-playing dataset]
Training
Using LLaMA-Factory to do the fine-tuning process. Here are the parameters: (TODO)
Use the following method to quickly call the GLM-4-9B-Chat language model
Use the transformers backend for inference:
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer
device = "cuda"
tokenizer = AutoTokenizer.from_pretrained("derek33125/project-angel-chatglm4", trust_remote_code=True)
query = "我感到很悲伤"
inputs = tokenizer.apply_chat_template([{"role": "user", "content": query}],
add_generation_prompt=True,
tokenize=True,
return_tensors="pt",
return_dict=True
)
inputs = inputs.to(device)
model = AutoModelForCausalLM.from_pretrained(
"derek33125/project-angel-chatglm4",
torch_dtype=torch.bfloat16,
low_cpu_mem_usage=True,
trust_remote_code=True
).to(device).eval()
gen_kwargs = {"max_length": 2500, "do_sample": True, "top_k": 1}
with torch.no_grad():
outputs = model.generate(**inputs, **gen_kwargs)
outputs = outputs[:, inputs['input_ids'].shape[1]:]
print(tokenizer.decode(outputs[0], skip_special_tokens=True))