Upload dpmpp-v2.patch

patch/dpmpp-v2.patch

@@ -0,0 +1,75 @@
+From 4394a62004260c3b9d781488e85f959a70910af1 Mon Sep 17 00:00:00 2001
+From: Andrew Powers-Holmes <[email protected]>
+Date: Sat, 8 Apr 2023 15:11:43 +1000
+Subject: [PATCH] add DPMPP 2M V2
+
+---
+ modules/sd_samplers_kdiffusion.py | 16 +++++++++-------
+ 1 file changed, 9 insertions(+), 7 deletions(-)
+
+diff --git a/modules/sd_samplers_kdiffusion.py b/modules/sd_samplers_kdiffusion.py
+index 93f0e55a..9202f4d4 100644
+--- a/modules/sd_samplers_kdiffusion.py
++++ b/modules/sd_samplers_kdiffusion.py
+@@ -27,12 +27,12 @@ samplers_k_diffusion = [
+     ('DPM2 a Karras', 'sample_dpm_2_ancestral', ['k_dpm_2_a_ka'], {'scheduler': 'karras', 'discard_next_to_last_sigma': True}),
+     ('DPM++ 2S a Karras', 'sample_dpmpp_2s_ancestral', ['k_dpmpp_2s_a_ka'], {'scheduler': 'karras'}),
+     ('DPM++ 2M Karras', 'sample_dpmpp_2m', ['k_dpmpp_2m_ka'], {'scheduler': 'karras'}),
++    ('DPM++ 2M v2', 'sample_dpmpp_2m_v2', ['k_dpmpp_2m'], {}),
++    ('DPM++ 2M Karras v2', 'sample_dpmpp_2m_v2', ['k_dpmpp_2m_ka'], {'scheduler': 'karras'}),
+     ('DPM++ SDE Karras', 'sample_dpmpp_sde', ['k_dpmpp_sde_ka'], {'scheduler': 'karras'}),
+ ]
+
+--
+
+---
+ k_diffusion/sampling.py | 36 ++++++++++++++++++++++++++++++++++++
+ 1 file changed, 36 insertions(+)
+
+diff --git a/repositories/k-diffusion/k_diffusion/sampling.py b/repositories/k-diffusion/k_diffusion/sampling.py
+index f050f88..1b0b282 100644
+--- a/repositories/k-diffusion/k_diffusion/sampling.py
++++ b/repositories/k-diffusion/k_diffusion/sampling.py
+@@ -605,4 +605,39 @@ def sample_dpmpp_2m(model, x, sigmas, extra_args=None, callback=None, disable=No
+             x = (sigma_fn(t_next) / sigma_fn(t)) * x - (-h).expm1() * denoised_d
+         old_denoised = denoised
+     return x
++
++
+[email protected]_grad()
++def sample_dpmpp_2m_v2(model, x, sigmas, extra_args=None, callback=None, disable=None):
++    """DPM-Solver++(2M)V2."""
++    extra_args = {} if extra_args is None else extra_args
++    s_in = x.new_ones([x.shape[0]])
++    sigma_fn = lambda t: t.neg().exp()
++    t_fn = lambda sigma: sigma.log().neg()
++    old_denoised = None
++
++    for i in trange(len(sigmas) - 1, disable=disable):
++        denoised = model(x, sigmas[i] * s_in, **extra_args)
++        if callback is not None:
++            callback({'x': x, 'i': i, 'sigma': sigmas[i], 'sigma_hat': sigmas[i], 'denoised': denoised})
++        t, t_next = t_fn(sigmas[i]), t_fn(sigmas[i + 1])
++        h = t_next - t
++
++        t_min = min(sigma_fn(t_next), sigma_fn(t))
++        t_max = max(sigma_fn(t_next), sigma_fn(t))
++
++        if old_denoised is None or sigmas[i + 1] == 0:
++            x = (t_min / t_max) * x - (-h).expm1() * denoised
++        else:
++            h_last = t - t_fn(sigmas[i - 1])
++
++            h_min = min(h_last, h)
++            h_max = max(h_last, h)
++            r = h_max / h_min
++
++            h_d = (h_max + h_min) / 2
++            denoised_d = (1 + 1 / (2 * r)) * denoised - (1 / (2 * r)) * old_denoised
++            x = (t_min / t_max) * x - (-h_d).expm1() * denoised_d
++
++        old_denoised = denoised
++    return x
+--
+2.34.1
+