DCLM-7B-Chat

This is a fine-tuned version of the DCLM-7B baseline model trained for chat completions.

Quick start

To use the model, open_lm must first be installed:

pip install git+https://github.com/mlfoundations/open_lm.git

Then simply load the model and generate responses:

from open_lm.hf import *
from transformers import (
    AutoModelForCausalLM,
    AutoTokenizer,
)


model = AutoModelForCausalLM.from_pretrained("mathewhe/DCLM-7B-Chat")
tokenizer = AutoTokenizer.from_pretrained("mathewhe/DCLM-7B-Chat")

messages = [
    {"role": "user", "content": "What is an LLM?"},
]

inputs = tokenizer.apply_chat_template(messages)

print(tokenizer.decode(model.generate(**inputs)[0]))

Alternatively, copy the included chat_class.py module into your local directory and just import the Chat class:

from chat_class import Chat
chat = Chat()  # default args: Chat("mathewhe/DCLM-7B-Chat", device="cuda")

# for one-off instructions
instruction = "Write a list of ingredients for banana pudding."
print(chat.instruct(instruction))

# for multi-turn chat
response1 = chat.message("Who was Stan Lee?")
response2 = chat.message("What was his wife's name?")

# to reset the chat
chat.reset()

Chat template

This model uses the following chat template and does not support a separate system prompt:

<|endoftext|>[INST] <user-message> [/INST][ASST] <llm-response> [/ASST]<|endoftext|>

The included tokenizer will correctly format messages, so you should not have to manually format the input text.

Instead, use the tokenizer's apply_chat_template() method on a list of messages. Each message should be a dict with two keys:

• "role": Either "user" or "assistant".
• "content": The message to include.

For example:

from transformers import AutoTokenizer

tokenizer = AutoTokenizer.from_pretrained("mathewhe/DCLM-7B-Chat")

messages = [
    {"role": "user", "content": "Solve for x: 3x=4"},
    {"role": "assistant", "content": "3x=4\n(3x)/3=(4)/3\nx=4/3"},
    {"role": "user", "content": "Please explain your work."},
]
print(tokenizer.apply_chat_template(messages, tokenize=False)

outputs

<|endoftext|>[INST] Solve for x: 3x=4 [/INST][ASST] 3x=4
(3x)/3=(4)/3
x=4/3 [/ASST]<|endoftext|><|endoftext|>[INST] Please explain your work [/INST]

See the example code in the included chat_class.py module for more details.