vllm_omni.model_executor.models.minimax_music3.sampling ¶

Deterministic guided sampling for MiniMax Music 3.

Two things happen per code, in this order:

Classifier-free guidance. Every request decodes twice, once on the real prompt and once on a prompt whose caption and lyrics have been replaced by <|audio_cfg|>. Codes are drawn from uncond + (cond - uncond) * 1.5. For c0 the candidate set is additionally masked to the conditioned row's own top 50, so the guided draw cannot wander outside what the conditioned branch already considered plausible.
Seeded top-k. A batched, order-independent categorical draw: each row gets its own RNG stream keyed by (seed, position), so a request's audio depends only on its own seed and never on who else happens to share the batch.

The RNG is a MurmurHash3 x86_32 of (seed_lo, seed_hi, position, column) turned into Gumbel noise, and the draw is an argmax over logits + gumbel. That is mathematically a categorical sample from softmax(logits), but unlike torch.multinomial it needs no per-row generator state, so it is CUDA-graph safe and reproducible across batch compositions.

apply_c0_cfg ¶

apply_c0_cfg(
    cond: Tensor,
    uncond: Tensor,
    *,
    scale: float = AR_CFG_SCALE,
) -> Tensor

Guide cond by uncond, restricted to cond's own top candidates.

The mask is a second, separate restriction from the top-k the sampler applies afterwards. Dropping it widens the distribution the model draws from and audibly loosens how closely the output follows the caption.

apply_depth_cfg ¶

apply_depth_cfg(
    cond: Tensor,
    uncond: Tensor,
    *,
    scale: float = AR_CFG_SCALE,
) -> Tensor

Guide the depth heads. Unlike c0 these are not additionally masked.

draw_from_gumbel ¶

draw_from_gumbel(
    logits: Tensor, gumbel: Tensor, *, top_k: int = AR_TOP_K
) -> Tensor

Argmax of top_k-masked logits plus precomputed Gumbel noise.

Parameters:

Name	Type	Description	Default
`logits`	`Tensor`	`[rows, vocab]` unnormalized scores.	required
`gumbel`	`Tensor`	`[rows, vocab]` float64 noise from :func:`gumbel_noise`.	required
`top_k`	`int`	Candidates retained per row before the draw.	`AR_TOP_K`

Returns:

Type	Description
`Tensor`	`[rows]` sampled column indices.

gumbel_noise ¶

gumbel_noise(
    seeds: Tensor, positions: Tensor, num_columns: int
) -> Tensor

Return the Gumbel noise for one draw per position, [..., cols].

Parameters:

Name	Type	Description	Default
`seeds`	`Tensor`	`[rows]` per-request seeds.	required
`positions`	`Tensor`	`[rows]` or `[rows, draws]` draw identifiers.	required
`num_columns`	`int`	Candidate count, i.e. the width of the logits.	required

Returns:

Type	Description
`Tensor`	float64 noise shaped `positions.shape + (num_columns,)`.

murmur_hash32 ¶

murmur_hash32(
    seeds: Tensor, positions: Tensor, columns: Tensor
) -> Tensor

Hash (seed, position, column) to a uint32 held in int64.

The 64-bit seed is consumed as two 32-bit blocks, then the position and the column, for a hashed length of 16 bytes.

Parameters:

Name	Type	Description	Default
`seeds`	`Tensor`	`[rows]` per-request seeds.	required
`positions`	`Tensor`	`[rows]` draw identifiers, unique per (frame, codebook).	required
`columns`	`Tensor`	`[cols]` candidate indices, which must be `arange(cols)`.	required

Returns:

Type	Description
`Tensor`	`[rows, cols]` int64 tensor holding uint32 values.

row_hash_state ¶

row_hash_state(seeds: Tensor, positions: Tensor) -> Tensor

Hash state after the seed and position blocks, before the columns.

Parameters:

Name	Type	Description	Default
`seeds`	`Tensor`	`[rows]` per-request seeds.	required
`positions`	`Tensor`	`[rows]` or `[rows, draws]` draw identifiers. The extra dimension lets every codebook of a frame be hashed in one batch instead of one dispatch per codebook; the arithmetic is elementwise integer, so a batched call is bit-identical to the loop it replaces.	required

Returns:

Type	Description
`Tensor`	A tensor shaped like `positions`, int64 holding uint32.

sample_topk_seeded ¶

sample_topk_seeded(
    logits: Tensor,
    seeds: Tensor,
    positions: Tensor,
    *,
    top_k: int = AR_TOP_K,
) -> Tensor

Draw one column per row from its own top-k, under its own RNG stream.

Parameters:

Name	Type	Description	Default
`logits`	`Tensor`	`[rows, vocab]` unnormalized scores.	required
`seeds`	`Tensor`	`[rows]` per-request sampling seeds.	required
`positions`	`Tensor`	`[rows]` draw identifiers, unique per (frame, codebook).	required
`top_k`	`int`	Candidates retained per row before the draw.	`AR_TOP_K`

Returns:

Type	Description
`Tensor`	`[rows]` sampled column indices.