nodes/billwuhao-ComfyUI_IndexTTS
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

fix high version transformers, shape '[1, 1815]' is invalid for input of size 24582 error

Browse Source
This commit is contained in:
billwuhao
2025-05-09 03:25:27 +08:00
parent 86245896e8
commit c22ea93124
3 changed files with 12 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files
indextts/gpt/model.py
+10 -6
View File
@@ -3,7 +3,7 @@ import functools
import torch
import torch.nn as nn
import torch.nn.functional as F
from transformers import GPT2Config, GPT2PreTrainedModel, LogitsProcessorList
from transformers import GPT2Config, GPT2LMHeadModel, GPT2Model, GPT2PreTrainedModel, LogitsProcessorList
from transformers.modeling_outputs import CausalLMOutputWithCrossAttentions
from transformers.utils.model_parallel_utils import (assert_device_map,
get_device_map)
@@ -37,7 +37,7 @@ class ResBlock(nn.Module):
return F.relu(self.net(x) + x)
class GPT2InferenceModel(GPT2PreTrainedModel):
class GPT2InferenceModel(GPT2LMHeadModel):
def __init__(self, config, gpt, text_pos_emb, embeddings, norm, linear, kv_cache=False):
super().__init__(config)
self.transformer = gpt
@@ -254,7 +254,6 @@ def build_hf_gpt_transformer(layers, model_dim, heads, max_mel_seq_len, max_text
"""
GPT-2 implemented by the HuggingFace library.
"""
from transformers import GPT2Config, GPT2Model
gpt_config = GPT2Config(vocab_size=256, # Unused.
n_positions=max_mel_seq_len + max_text_seq_len,
n_ctx=max_mel_seq_len + max_text_seq_len,
@@ -388,7 +387,7 @@ class UnifiedVoice(nn.Module):
def post_init_gpt2_config(self, use_deepspeed=False, kv_cache=False, half=False):
seq_length = self.max_mel_tokens + self.max_text_tokens + 2
gpt_config = GPT2Config(
vocab_size=self.max_mel_tokens,
vocab_size=self.number_mel_codes,
n_positions=seq_length,
n_ctx=seq_length,
n_embd=self.model_dim,
@@ -396,6 +395,7 @@ class UnifiedVoice(nn.Module):
n_head=self.heads,
gradient_checkpointing=False,
use_cache=True,
# attn_implementation="flash_attention_2"
)
self.inference_model = GPT2InferenceModel(
gpt_config,
@@ -410,14 +410,18 @@ class UnifiedVoice(nn.Module):
import deepspeed
self.ds_engine = deepspeed.init_inference(model=self.inference_model,
mp_size=1,
replace_with_kernel_inject=False,
use_triton=False,
triton_autotune=False,
replace_with_kernel_inject=True,
dtype=torch.float16)
self.inference_model = self.ds_engine.module.eval()
elif use_deepspeed and torch.cuda.is_available():
import deepspeed
self.ds_engine = deepspeed.init_inference(model=self.inference_model,
mp_size=1,
replace_with_kernel_inject=False,
use_triton=False,
triton_autotune=False,
replace_with_kernel_inject=True,
dtype=torch.float32)
self.inference_model = self.ds_engine.module.eval()
else:
indexttsnode.py
+1 -1
View File
@@ -513,7 +513,7 @@ class IndexTTSRun:
"top_k": ("INT", {"default": 30, "min": 0, "max": 1000, "step": 1}),
"top_p": ("FLOAT", {"default": 0.8, "min": 0.0, "max": 1.0, "step": 0.01}),
"temperature": ("FLOAT", {"default": 1.0, "min": 0.0, "max": 2.0, "step": 0.01}),
"max_mel_tokens": ("INT", {"default": 600, "min": 0, "max": 100000, "step": 1}),
"max_mel_tokens": ("INT", {"default": 1000, "min": 0, "max": 100000, "step": 1}),
"bucket_enable": ("BOOLEAN", {"default": True}),
"fast_inference": ("BOOLEAN", {"default": True}),
"unload_model": ("BOOLEAN", {"default": True}),
pyproject.toml
+1 -1
View File
@@ -1,7 +1,7 @@
[project]
name = "indextts-mw"
description = "IndexTTS Voice Cloning Nodes for ComfyUI. High-quality voice cloning, very fast, supports Chinese and English, and allows custom voice styles."
version = "1.0.1"
version = "1.0.2"
license = {file = "LICENSE"}
dependencies = ["# accelerate==0.25.0", "# transformers==4.36.2", "# tokenizers==0.15.0", "# cn2an==0.5.22", "# ffmpeg-python==0.2.0", "# Cython==3.0.7", "# g2p-en==2.1.0", "# jieba==0.42.1", "# keras==2.9.0", "# numba==0.58.1", "# numpy==1.26.2", "# pandas==2.1.3", "# matplotlib==3.8.2", "# opencv-python==4.9.0.80", "# vocos==0.1.0", "# accelerate==0.25.0", "# tensorboard==2.9.1", "omegaconf", "sentencepiece", "librosa", "tqdm", "# deepspeeds # Use it to accelerate model inference"]