Changing a single example affects forward pass for other examples in a batch

[Mayank Mishra]

@stas00 , I wrote this script to do get the conditional NLL for the labels given the context.
Tried different batches with only the first example changing and rest of the examples fixed in the batch. However, after a certain point, the changing of first examples, affects the NLL for other examples.

This is not supposed to happen.

import torch
from transformers import AutoTokenizer, AutoModelForCausalLM

model_name = "bigscience/bloom"

tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(
    model_name,
    device_map="auto",
    max_memory={0: '0GIB', 1: '51GIB', 2: '51GIB', 3: '51GIB',
                4: '51GIB', 5: '51GIB', 6: '51GIB', 7: '51GIB'},
    torch_dtype=torch.bfloat16,
)

model.eval()

def compute_gen_loss(lm_logits: torch.Tensor, labels: torch.Tensor) -> torch.Tensor:
    batch_size = labels.shape[0]
    shift_logits = lm_logits[..., :-1, :].contiguous()
    shift_labels = labels[..., 1:].contiguous()

    loss_fct = torch.nn.CrossEntropyLoss(reduction="none")
    loss = loss_fct(
        shift_logits.view(-1, shift_logits.size(-1)),
        shift_labels.view(-1)
    )
    loss = loss.reshape(batch_size, -1)
    loss = loss.sum(dim=-1) / (shift_labels != -100).sum(dim=-1)
    return loss


def pad_ids(arrays, padding, max_length=-1):
    if (max_length < 0):
        max_length = max(list(map(len, arrays)))

    arrays = [[padding] * (max_length - len(array)) +
              array for array in arrays]

    return arrays


def forward(text: list, labels: str, conditional: bool = True):
    input_tokens = tokenizer(text).input_ids
    label_tokens = tokenizer(labels).input_ids

    input_ids = [x + y for (x, y) in zip(input_tokens, label_tokens)]
    attention_mask = [(len(x) + len(y)) * [1]
                      for (x, y) in zip(input_tokens, label_tokens)]
    if (conditional):
        labels = [[-100] * len(x) + y for (x, y)
                  in zip(input_tokens, label_tokens)]
    else:
        labels = input_ids

    pad = 3
    input_ids = pad_ids(input_ids, pad)
    attention_mask = pad_ids(attention_mask, 0)
    # labels need to be on output device
    labels = pad_ids(labels, -100)

    input_ids = torch.tensor(input_ids)
    attention_mask = torch.tensor(attention_mask)
    labels = torch.tensor(labels)
    lm_logits = model(
        input_ids=input_ids,
        attention_mask=attention_mask
    ).logits

    print(compute_gen_loss(lm_logits, labels).cpu().tolist())

text = [
    "DeepSpeed",
    "DeepSpeed is a",
    "DeepSpeed is a machine",
    "DeepSpeed is a machine learning framework",
]
labels = [
    " is awesome.",
    " good person.",
    " that can wipe out the planet.",
    " for generating memes.",
]
forward(text, labels)

labels[0] = " is awesome. really awesome"
forward(text, labels)

labels[0] = " is awesome. really awesome. Try it."
forward(text, labels)

labels[0] = " is awesome. really awesome. Try it. You'll be surprised"
forward(text, labels)

labels[0] = " is awesome. really awesome. Try it. You'll be surprised. BLOOM was trained using DeepSpeed."
forward(text, labels)

labels[0] = " is awesome. really awesome. Try it. You'll be surprised. BLOOM was trained using DeepSpeed. Oh no the values are bugging out now."
forward(text, labels)

Output:

[4.8125, 5.1875, 3.296875, 5.09375]
[5.625, 5.1875, 3.296875, 5.09375]
[4.375, 5.1875, 3.296875, 5.09375]
[4.0625, 5.1875, 3.28125, 5.09375]
[3.953125, 5.1875, 3.28125, 5.0625]
[4.25, 5.1875, 3.296875, 5.09375]

Value drops from 3.29 to 3.28 in column 2 when only example for column 0 is changed. Even column 3 changes in last case.
Only column 0 is supposed to change here.

[Stas Bekman]

@mayank31398, I think it'd be much better to file this issue with transformers, since this code isn't Mega-DS-related.

I know several folks have been tweaking the bloom modeling code a lot recently, so you may want to tag them on that (peek into the history of https://github.com/huggingface/transformers/blame/main/src/transformers/models/bloom/modeling_bloom.py)

[Mayank Mishra]

Thanks, I will do that.
Closing this issue here.

[Mayank Mishra]

Filed this issue huggingface/transformers#18809

[Stas Bekman]

FYI: I have re-tagged that new issue to those who have been actively tweaking the model, so they are the best to talk to.