The original idea was to use this 1.8B model, divide the dataset based on task specific capabilities, train models and transform them into a mixture of experts.
Hyperparameters: adamw with eps of 1e-8, cosine decay w/ 20% warmup, lr=2e-5.
Format:
<|system|></s><|prompt|></s><|answer|>
Benchamrks:
WIP
Example:
from transformers import AutoModelForCausalLM, AutoTokenizer, TextStreamer, StoppingCriteria
import torch
class MyStoppingCriteria(StoppingCriteria):
def __init__(self, target_sequence, prompt):
self.target_sequence = target_sequence
self.prompt=prompt
def __call__(self, input_ids, scores, **kwargs):
generated_text = tokenizer.decode(input_ids[0])
generated_text = generated_text.replace(self.prompt,'')
if self.target_sequence in generated_text:
return True
return False
def __len__(self):
return 1
def __iter__(self):
yield self
modelpath="aloobun/Cypher-CoT-1.8B"
model = AutoModelForCausalLM.from_pretrained(
modelpath,
torch_dtype=torch.bfloat16,
device_map="cuda",
trust_remote_code=True,
)
tokenizer = AutoTokenizer.from_pretrained(
modelpath,
trust_remote_code=True,
use_fast=False,
)
prompt = "<|prompt|>James takes a spinning class 3 times a week. He works out for 1.5 hours each class and burns 7 calories per minute. How many calories does he burn per week?</s><|answer|>"
encoded_input = tokenizer(prompt, return_tensors='pt')
input_ids=encoded_input['input_ids'].cuda()
streamer = TextStreamer(tokenizer=tokenizer, skip_prompt=True)
op = model.generate(
input_ids,
streamer=streamer,
pad_token_id=tokenizer.eos_token_id,
do_sample=True,
temperature=0.7,
top_p=0.8,
max_new_tokens=512,
stopping_criteria=MyStoppingCriteria("</s>", prompt)
)
Output:
James takes a spinning class 3 times a week, so he spends a total of 3 * 1.5 = 4.5 hours in the class each week.
Since there are 60 minutes in an hour, this is equivalent to 4.5 * 60 = 270 minutes.
If he burns 7 calories per minute, then he burns a total of 270 * 7 = 1890 calories per week.
####1890
The answer is: 1890
