From 9e27d2e02e66cce2cd4ad1db37304f387fd30ec8 Mon Sep 17 00:00:00 2001
From: mrq <mrq@ecker.tech>
Date: Wed, 16 Apr 2025 15:25:45 -0500
Subject: [PATCH] huggingface zerogpu cringe

---
 vall_e/emb/codecs/__init__.py |  0
 vall_e/emb/transcribe.py      |  7 ++++++-
 vall_e/models/base_v2.py      | 13 -------------
 3 files changed, 6 insertions(+), 14 deletions(-)
 create mode 100644 vall_e/emb/codecs/__init__.py

diff --git a/vall_e/emb/codecs/__init__.py b/vall_e/emb/codecs/__init__.py
new file mode 100644
index 0000000..e69de29
diff --git a/vall_e/emb/transcribe.py b/vall_e/emb/transcribe.py
index e496eb0..d894ec5 100644
--- a/vall_e/emb/transcribe.py
+++ b/vall_e/emb/transcribe.py
@@ -18,7 +18,12 @@ except Exception as e:
 	pass
 """
 
-from transformers import pipeline
+try:
+	from transformers import pipeline
+except Exception as e:
+	def _kludge_cringe():
+		raise e
+	pipeline = _kludge_cringe
 
 from functools import cache
 from tqdm.auto import tqdm
diff --git a/vall_e/models/base_v2.py b/vall_e/models/base_v2.py
index 66d387e..710260d 100644
--- a/vall_e/models/base_v2.py
+++ b/vall_e/models/base_v2.py
@@ -123,16 +123,6 @@ class FiniteAudioEncoder(nn.Module):
 		self.level_weights = nn.Parameter(torch.ones(n_levels) / math.sqrt(n_levels)) if use_level_weights else None
 		self.use_ffn = use_ffn
 
-		if use_ffn:
-			nn.init.xavier_uniform_(self.proj[0].weight)
-			nn.init.xavier_uniform_(self.proj[2].weight)
-
-			nn.init.zeros_(self.proj[0].bias)
-			nn.init.zeros_(self.proj[2].bias)
-		elif token_dim != d_model:
-			nn.init.xavier_uniform_(self.proj.weight)
-			nn.init.zeros_(self.proj.bias)
-
 	def forward(self, xi: Tensor, dropout_mask = None, dropout_token = None, stability = None, mode = None ) -> Tensor:
 		# empty
 		if xi.shape[0] == 0:
@@ -162,9 +152,6 @@ class FiniteAudioEncoder(nn.Module):
 
 		if self.level_weights is None:
 			x = x.sum(dim=1)
-		elif stability:
-			weights = F.softmax(self.level_weights.float(), dim=0).view(1, -1, 1)
-			x = (x.float() * weights).sum(dim=1).to(xi.dtype)
 		else:
 			weights = F.softmax(self.level_weights, dim=0).view(1, -1, 1)
 			x = (x * weights).sum(dim=1)