[BUG] The result of SUMBT model

[zyds]

Hi! When I tried to evaluate the translation training SUMBT model, I found that the eval mode was not set, which had a certain impact on the results. According to the results of my local test, I found that there is a difference of two points on the MutliWOZ-zh human-val dataset. I think it may be necessary to re-evaluate the SUMBT model after modifying the code. The current results
are not real model performance.

My Local Result on MultiWOZ-zh
not set eval mode
{'Joint Acc': 0.4821722435545804, 'Turn Acc': 0.9738983360760534, 'Joint F1': 0.8826705748001639}
set eval mode
{'Joint Acc': 0.49972572682391664, 'Turn Acc': 0.9751935149631128, 'Joint F1': 0.8885012208542876}

[zyds]

def reformat_state(state):
    if 'belief_state' in state:
        state = state['belief_state']
    new_state = []
    for domain in state.keys():
        domain_data = state[domain]
        if 'semi' in domain_data:
            domain_data = domain_data['semi']
            for slot in domain_data.keys():
                val = domain_data[slot]
                if val is not None and val not in ['', 'not mentioned', '未提及', '未提到', '没有提到']:
                    new_state.append(domain + '-' + slot + '-' + val)
    # lower
    new_state = [item.lower() for item in new_state]
    return new_state

This code is in dst/evaluate.py, I want to know about dialog state, can it be calculated using only the semi part?

[zqwerty]

@function2-llx please look at this

[zqwerty]

Hi! When I tried to evaluate the translation training SUMBT model, I found that the eval mode was not set, which had a certain impact on the results. According to the results of my local test, I found that there is a difference of two points on the MutliWOZ-zh human-val dataset. I think it may be necessary to re-evaluate the SUMBT model after modifying the code. The current results
are not real model performance.

My Local Result on MultiWOZ-zh
not set eval mode
{'Joint Acc': 0.4821722435545804, 'Turn Acc': 0.9738983360760534, 'Joint F1': 0.8826705748001639}
set eval mode
{'Joint Acc': 0.49972572682391664, 'Turn Acc': 0.9751935149631128, 'Joint F1': 0.8885012208542876}

How do you get the above result? by running dst/evaluate.py ?

[zyds]

Hi! When I tried to evaluate the translation training SUMBT model, I found that the eval mode was not set, which had a certain impact on the results. According to the results of my local test, I found that there is a difference of two points on the MutliWOZ-zh human-val dataset. I think it may be necessary to re-evaluate the SUMBT model after modifying the code. The current results
are not real model performance.
My Local Result on MultiWOZ-zh
not set eval mode
{'Joint Acc': 0.4821722435545804, 'Turn Acc': 0.9738983360760534, 'Joint F1': 0.8826705748001639}
set eval mode
{'Joint Acc': 0.49972572682391664, 'Turn Acc': 0.9751935149631128, 'Joint F1': 0.8885012208542876}

How do you get the above result? by running dst/evaluate.py ?

Yes, but the results I report are not using the pre training model provided by the project. However, using the pre training model provided by the project, I also got similar results.

[zqwerty]

update SUMBT & test results #69