Skip to content

[JAX] Replace jnp.clip(..., a_min=..., a_max=...) with jnp.clip(..., min=..., max=...). #41

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
copybara-service merged 1 commit into from
Mar 27, 2026
+3 −3
Merged

[JAX] Replace jnp.clip(..., a_min=..., a_max=...) with jnp.clip(..., min=..., max=...). #41

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion hackable_diffusion/lib/loss/discrete.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -184,7 +184,7 @@ def _weight_fn(
alpha_der = utils.egrad(schedule.alpha)(time)
alpha = utils.flatten_non_batch_dims(alpha)
alpha_der = utils.flatten_non_batch_dims(alpha_der)
weight = -1.0 * alpha_der / jnp.clip(1.0 - alpha, a_min=1e-12)
weight = -1.0 * alpha_der / jnp.clip(1.0 - alpha, min=1e-12)
return weight

return compute_discrete_diffusion_loss(
Expand Down
4 changes: 2 additions & 2 deletions hackable_diffusion/lib/sampling/discrete_step_sampler.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -708,7 +708,7 @@ def update(
# (bsz, *seq_len, M)

# Calculate integration weights: W(x_0) = p(x_0 | x_t) / q(x_t | x_0).
w_x0 = p_x0 / jnp.clip(q_xt_given_x0, a_min=1e-12)
w_x0 = p_x0 / jnp.clip(q_xt_given_x0, min=1e-12)
# (bsz, *seq_len, M)
sum_w = jnp.sum(w_x0, axis=-1, keepdims=True)
# (bsz, *seq_len, 1)
Expand All @@ -722,7 +722,7 @@ def update(
# (bsz, *seq_len, M)

# Convert back to logits for safe categorical sampling
total_logit = jnp.log(jnp.clip(p_xs, a_min=1e-12))
total_logit = jnp.log(jnp.clip(p_xs, min=1e-12))

# Sample and format the new state
new_xt = jax.random.categorical(key=key, logits=total_logit)[..., None]
Expand Down
Loading