From 81c176ea1b00336b5490ac636f49073a71d464c6 Mon Sep 17 00:00:00 2001
From: lalitpagaria <pagaria.lalit@gmail.com>
Date: Sun, 15 Nov 2020 18:03:57 +0100
Subject: [PATCH 1/3] Removing device information from generator model
arguments as it is handled by itself.
---
haystack/generator/transformers.py | 5 ++---
1 file changed, 2 insertions(+), 3 deletions(-)
diff --git a/haystack/generator/transformers.py b/haystack/generator/transformers.py
index a9d055b167..983ba465f2 100644
--- a/haystack/generator/transformers.py
+++ b/haystack/generator/transformers.py
@@ -153,8 +153,7 @@ def _get_contextualized_inputs(self, texts: List[str], question: str, titles: Op
truncation=True,
)
- return contextualized_inputs["input_ids"].to(self.device), \
- contextualized_inputs["attention_mask"].to(self.device)
+ return contextualized_inputs["input_ids"], contextualized_inputs["attention_mask"]
def _prepare_passage_embeddings(self, docs: List[Document], embeddings: List[Optional[numpy.ndarray]]) -> torch.Tensor:
@@ -170,7 +169,7 @@ def _prepare_passage_embeddings(self, docs: List[Document], embeddings: List[Opt
embeddings_in_tensor = torch.cat(
[torch.from_numpy(embedding).unsqueeze(0) for embedding in embeddings],
dim=0
- ).to(self.device)
+ )
return embeddings_in_tensor
From 0c3f80c6a173d3d69578956a628e2451d5d43b25 Mon Sep 17 00:00:00 2001
From: lalitpagaria <pagaria.lalit@gmail.com>
Date: Sun, 15 Nov 2020 18:18:33 +0100
Subject: [PATCH 2/3] num_return_sequences of should not be greate than
num_beams
---
haystack/generator/transformers.py | 12 +++++++++++-
1 file changed, 11 insertions(+), 1 deletion(-)
diff --git a/haystack/generator/transformers.py b/haystack/generator/transformers.py
index 983ba465f2..fa8ff2a9bc 100644
--- a/haystack/generator/transformers.py
+++ b/haystack/generator/transformers.py
@@ -90,7 +90,6 @@ def __init__(
"""
self.model_name_or_path = model_name_or_path
- self.top_k_answers = top_k_answers
self.max_length = max_length
self.min_length = min_length
self.generator_type = generator_type
@@ -99,6 +98,12 @@ def __init__(
self.prefix = prefix
self.retriever = retriever
+ if top_k_answers > self.num_beams:
+ top_k_answers = self.num_beams
+ logger.warning(f'top_k_answers value should not be greater than num_beams, hence setting it to {num_beams}')
+
+ self.top_k_answers = top_k_answers
+
if use_gpu and torch.cuda.is_available():
self.device = torch.device("cuda")
else:
@@ -201,6 +206,11 @@ def predict(self, question: str, documents: List[Document], top_k: Optional[int]
top_k_answers = top_k if top_k is not None else self.top_k_answers
+ if top_k_answers > self.num_beams:
+ top_k_answers = self.num_beams
+ logger.warning(f'top_k_answers value should not be greater than num_beams, '
+ f'hence setting it to {top_k_answers}')
+
# Flatten the documents so easy to reference
flat_docs_dict: Dict[str, Any] = {}
for document in documents:
From bf18ba577206311964bc3e5ad4e856fddcefb303 Mon Sep 17 00:00:00 2001
From: lalitpagaria <pagaria.lalit@gmail.com>
Date: Mon, 16 Nov 2020 09:16:28 +0100
Subject: [PATCH 3/3] Raise error when user use generator with GPU as currently
it is not supported
---
haystack/generator/transformers.py | 6 ++++--
1 file changed, 4 insertions(+), 2 deletions(-)
diff --git a/haystack/generator/transformers.py b/haystack/generator/transformers.py
index fa8ff2a9bc..c93fbc89c8 100644
--- a/haystack/generator/transformers.py
+++ b/haystack/generator/transformers.py
@@ -106,6 +106,7 @@ def __init__(
if use_gpu and torch.cuda.is_available():
self.device = torch.device("cuda")
+ raise AttributeError("Currently RAGenerator does not support GPU, try with use_gpu=False")
else:
self.device = torch.device("cpu")
@@ -158,7 +159,8 @@ def _get_contextualized_inputs(self, texts: List[str], question: str, titles: Op
truncation=True,
)
- return contextualized_inputs["input_ids"], contextualized_inputs["attention_mask"]
+ return contextualized_inputs["input_ids"].to(self.device), \
+ contextualized_inputs["attention_mask"].to(self.device)
def _prepare_passage_embeddings(self, docs: List[Document], embeddings: List[Optional[numpy.ndarray]]) -> torch.Tensor:
@@ -174,7 +176,7 @@ def _prepare_passage_embeddings(self, docs: List[Document], embeddings: List[Opt
embeddings_in_tensor = torch.cat(
[torch.from_numpy(embedding).unsqueeze(0) for embedding in embeddings],
dim=0
- )
+ ).to(self.device)
return embeddings_in_tensor