vzhong / e3 Goto Github PK

View Code? Open in Web Editor NEW

48.0 48.0 10.0 31 KB

Dockerized code for E3: Entailment-driven Extracting and Editing for Conversational Machine Reading.

Home Page: https://arxiv.org/abs/1906.05373

Dockerfile 1.15% Shell 1.03% Python 97.83%

deep-learning machine-learning nlp

e3's Introduction

👋 Hi, I’m @vzhong. I am an Assistant Professor at the University of Waterloo.
📫 Find me at https://victorzhong.com

e3's People

Contributors

Stargazers

Watchers

Forkers

codeaudit marcelomata lrh000 nishthaa yifan-gao zhangshuoyang fivekilometers ammieqi tonystark262 dwtcourses

e3's Issues

No module named 'embeddings'

Hi, I'm running the non-docker setup.
When I tried to run
python -c "import embeddings as e; e.GloveEmbedding()"
I got
ModuleNotFoundError: No module named 'embeddings'

Can you help me with that?

Difference between Equation 14 in paper and its implementation

In equation 14 of E3 paper, the summary vector C is computed based on the extracted spans (summation from s_i to e_i of each span).

However, the implementation here considers all tokens for the summary vector C:

e3/model/entail.py

Lines 59 to 61 in 0c6b771

 inp_attn_score = self.inp_attn_scorer(self.dropout(out['bert_enc'])).squeeze(2) - (1-out['input_mask'].float()).mul(1e20) 

 inp_attn = F.softmax(inp_attn_score, dim=1).unsqueeze(2).expand_as(out['bert_enc']).mul(self.dropout(out['bert_enc'])).sum(1) 

 out['clf_scores'] = self.class_clf(self.dropout(inp_attn))

Permission denied when running train_sharc.py via Docker

With both Dockerfile and wrap.sh unchanged, I was able to run docker/wrap.sh python preprocess_sharc.py
but encountered PermissionError: [Errno 13] Permission denied: '/opt/code/save/default-entail/train.log' when running NV_GPU=0 docker/wrap.sh python train_sharc.py
I think this was caused by the access permission to /opt/, is there any workaround?

preprocess_sharc.py: func filter_answer() and get_bullets() may not work as expected

When I run preprocess_sharc.py, I find that filter_answer never filter out any token:

e3/preprocess_sharc.py

Lines 48 to 49 in f9b5d1e

 def filter_answer(answer): 

 return detokenize([a for a in answer if a['orig'] not in MATCH_IGNORE])

Also, get_bullets will always return a empty list:

e3/preprocess_sharc.py

Lines 107 to 118 in f9b5d1e

 def get_bullets(context): 

 indices = [i for i, c in enumerate(context) if c == '*'] 

 pairs = list(zip(indices, indices[1:] + [len(context)])) 

 cleaned = [] 

 for s, e in pairs: 

 while not context[e-1].strip(): 

 e -= 1 

 while not context[s].strip() or context[s] == '*': 

 s += 1 

 if e - s > 2 and e - 2 < 45: 

 cleaned.append((s, e-1)) 

 return cleaned

I suspect the possible reason could be revtok.tokenize will return a tokenized list with added spaces like Hello, which make token['orig'] fail to match tokens in MATCH_IGNORE and *.

Replicate results

Hi Vector,
I have some problems on replicating experimental results. Because my turing GPU does not support CUDA8, I only tried training from scratch and loading your trained models. But both methods cannot replicate results on dev set.

Dev set	Micro	Macro	B1	B4
E3 paper	68.02	73.36	67.14	53.67
train from scratch	66.43	72.53	57.88	41.37
binary model in docker	66.43	72.53	57.1	40.68

vzhong / e3 Goto Github PK

e3's Introduction

e3's People

Contributors

Stargazers

Watchers

Forkers

e3's Issues

No module named 'embeddings'

Difference between Equation 14 in paper and its implementation

Permission denied when running train_sharc.py via Docker

preprocess_sharc.py: func filter_answer() and get_bullets() may not work as expected

Replicate results

Recommend Projects

React

Vue.js

Typescript

TensorFlow

Django

Laravel

D3

Recommend Topics

javascript

web

server

Machine learning

Visualization

Game

Recommend Org

Facebook

Microsoft

Google

Alibaba

D3

Tencent

	inp_attn_score = self.inp_attn_scorer(self.dropout(out['bert_enc'])).squeeze(2) - (1-out['input_mask'].float()).mul(1e20)
	inp_attn = F.softmax(inp_attn_score, dim=1).unsqueeze(2).expand_as(out['bert_enc']).mul(self.dropout(out['bert_enc'])).sum(1)
	out['clf_scores'] = self.class_clf(self.dropout(inp_attn))

	def filter_answer(answer):
	return detokenize([a for a in answer if a['orig'] not in MATCH_IGNORE])

	def get_bullets(context):
	indices = [i for i, c in enumerate(context) if c == '*']
	pairs = list(zip(indices, indices[1:] + [len(context)]))
	cleaned = []
	for s, e in pairs:
	while not context[e-1].strip():
	e -= 1
	while not context[s].strip() or context[s] == '*':
	s += 1
	if e - s > 2 and e - 2 < 45:
	cleaned.append((s, e-1))
	return cleaned