Mamba-790M

Mamba Models with hf_integration.

For modeling codes: mamba-hf

Usage:

from transformers import AutoModelForCausalLM , AutoTokenizer

model = AutoModelForCausalLM.from_pretrained('Q-bert/Mamba-790M', trust_remote_code=True)
tokenizer = AutoTokenizer.from_pretrained('Q-bert/Mamba-790M')

text = "Hi"

input_ids = tokenizer.encode(text, return_tensors="pt")

output = model.generate(input_ids, max_length=20, num_beams=5, no_repeat_ngram_size=2)

generated_text = tokenizer.decode(output[0], skip_special_tokens=True)

print(generated_text)

Hi, I'm looking for a new job. I've been working at a company for about a year now.

For Training:

from transformers import Trainer ,TrainingArguments
import torch
import os


class MambaTrainer(Trainer):
    def compute_loss(self, model, inputs, return_outputs=False):
        input_ids = inputs.pop("input_ids")
        lm_logits = model(input_ids)[0]

        labels = input_ids.to(lm_logits.device)
        shift_logits = lm_logits[:, :-1, :].contiguous()
        labels = labels[:, 1:].contiguous()

        loss_fct = torch.nn.CrossEntropyLoss()
        lm_loss = loss_fct(shift_logits.view(-1, shift_logits.size(-1)), labels.view(-1))

        return lm_loss

You must use this class for training. And fp16 must be False.

Credits:

https://huggingface.co./state-spaces

Special thanks to Albert Gu and Tri Dao for their articles. (https://arxiv.org/abs/2312.00752)