From 02d7abf5141431b9a3a8a189bb3136c71abd5e79 Mon Sep 17 00:00:00 2001 From: AUTOMATIC <16777216c@gmail.com> Date: Wed, 4 Jan 2023 12:35:07 +0300 Subject: helpful error message when trying to load 2.0 without config failing to load model weights from settings won't break generation for currently loaded model anymore --- modules/errors.py | 25 +++++++++++++++++++++++-- 1 file changed, 23 insertions(+), 2 deletions(-) (limited to 'modules/errors.py') diff --git a/modules/errors.py b/modules/errors.py index 372dc51a..a668c014 100644 --- a/modules/errors.py +++ b/modules/errors.py @@ -2,9 +2,30 @@ import sys import traceback +def print_error_explanation(message): + lines = message.strip().split("\n") + max_len = max([len(x) for x in lines]) + + print('=' * max_len, file=sys.stderr) + for line in lines: + print(line, file=sys.stderr) + print('=' * max_len, file=sys.stderr) + + +def display(e: Exception, task): + print(f"{task or 'error'}: {type(e).__name__}", file=sys.stderr) + print(traceback.format_exc(), file=sys.stderr) + + message = str(e) + if "copying a param with shape torch.Size([640, 1024]) from checkpoint, the shape in current model is torch.Size([640, 768])" in message: + print_error_explanation(""" +The most likely cause of this is you are trying to load Stable Diffusion 2.0 model without specifying its connfig file. +See https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Features#stable-diffusion-20 for how to solve this. + """) + + def run(code, task): try: code() except Exception as e: - print(f"{task}: {type(e).__name__}", file=sys.stderr) - print(traceback.format_exc(), file=sys.stderr) + display(task, e) -- cgit v1.2.3 From eb2223340cfdd58efaa157662c279fbf6c90c7d9 Mon Sep 17 00:00:00 2001 From: fuggy <45698918+nonetrix@users.noreply.github.com> Date: Mon, 16 Jan 2023 21:50:30 -0600 Subject: Fix typo --- modules/errors.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'modules/errors.py') diff --git a/modules/errors.py b/modules/errors.py index a668c014..a10e8708 100644 --- a/modules/errors.py +++ b/modules/errors.py @@ -19,7 +19,7 @@ def display(e: Exception, task): message = str(e) if "copying a param with shape torch.Size([640, 1024]) from checkpoint, the shape in current model is torch.Size([640, 768])" in message: print_error_explanation(""" -The most likely cause of this is you are trying to load Stable Diffusion 2.0 model without specifying its connfig file. +The most likely cause of this is you are trying to load Stable Diffusion 2.0 model without specifying its config file. See https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Features#stable-diffusion-20 for how to solve this. """) -- cgit v1.2.3 From 59146621e256269b85feb536edeb745da20daf68 Mon Sep 17 00:00:00 2001 From: AUTOMATIC <16777216c@gmail.com> Date: Mon, 23 Jan 2023 16:40:20 +0300 Subject: better support for xformers flash attention on older versions of torch --- modules/errors.py | 12 +++++++++++ modules/sd_hijack_optimizations.py | 42 ++++++++++++++++---------------------- 2 files changed, 30 insertions(+), 24 deletions(-) (limited to 'modules/errors.py') diff --git a/modules/errors.py b/modules/errors.py index a10e8708..f6b80dbb 100644 --- a/modules/errors.py +++ b/modules/errors.py @@ -24,6 +24,18 @@ See https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Features#stable """) +already_displayed = {} + + +def display_once(e: Exception, task): + if task in already_displayed: + return + + display(e, task) + + already_displayed[task] = 1 + + def run(code, task): try: code() diff --git a/modules/sd_hijack_optimizations.py b/modules/sd_hijack_optimizations.py index 9967359b..74452709 100644 --- a/modules/sd_hijack_optimizations.py +++ b/modules/sd_hijack_optimizations.py @@ -9,7 +9,7 @@ from torch import einsum from ldm.util import default from einops import rearrange -from modules import shared +from modules import shared, errors from modules.hypernetworks import hypernetwork from .sub_quadratic_attention import efficient_dot_product_attention @@ -279,6 +279,21 @@ def sub_quad_attention(q, k, v, q_chunk_size=1024, kv_chunk_size=None, kv_chunk_ ) +def get_xformers_flash_attention_op(q, k, v): + if not shared.cmd_opts.xformers_flash_attention: + return None + + try: + flash_attention_op = xformers.ops.MemoryEfficientAttentionFlashAttentionOp + fw, bw = flash_attention_op + if fw.supports(xformers.ops.fmha.Inputs(query=q, key=k, value=v, attn_bias=None)): + return flash_attention_op + except Exception as e: + errors.display_once(e, "enabling flash attention") + + return None + + def xformers_attention_forward(self, x, context=None, mask=None): h = self.heads q_in = self.to_q(x) @@ -291,18 +306,7 @@ def xformers_attention_forward(self, x, context=None, mask=None): q, k, v = map(lambda t: rearrange(t, 'b n (h d) -> b n h d', h=h), (q_in, k_in, v_in)) del q_in, k_in, v_in - if shared.cmd_opts.xformers_flash_attention: - op = xformers.ops.MemoryEfficientAttentionFlashAttentionOp - fw, bw = op - if not fw.supports(xformers.ops.fmha.Inputs(query=q, key=k, value=v, attn_bias=None)): - # print('xformers_attention_forward', q.shape, k.shape, v.shape) - # Flash Attention is not availabe for the input arguments. - # Fallback to default xFormers' backend. - op = None - else: - op = None - - out = xformers.ops.memory_efficient_attention(q, k, v, attn_bias=None, op=op) + out = xformers.ops.memory_efficient_attention(q, k, v, attn_bias=None, op=get_xformers_flash_attention_op(q, k, v)) out = rearrange(out, 'b n h d -> b n (h d)', h=h) return self.to_out(out) @@ -377,17 +381,7 @@ def xformers_attnblock_forward(self, x): q = q.contiguous() k = k.contiguous() v = v.contiguous() - if shared.cmd_opts.xformers_flash_attention: - op = xformers.ops.MemoryEfficientAttentionFlashAttentionOp - fw, bw = op - if not fw.supports(xformers.ops.fmha.Inputs(query=q, key=k, value=v)): - # print('xformers_attnblock_forward', q.shape, k.shape, v.shape) - # Flash Attention is not availabe for the input arguments. - # Fallback to default xFormers' backend. - op = None - else: - op = None - out = xformers.ops.memory_efficient_attention(q, k, v, op=op) + out = xformers.ops.memory_efficient_attention(q, k, v, op=get_xformers_flash_attention_op(q, k, v)) out = rearrange(out, 'b (h w) c -> b c h w', h=h) out = self.proj_out(out) return x + out -- cgit v1.2.3