optimum/habana/transformers/models/deepseek_v2/modeling_deepseek_v2.py [300:354]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
    def __init__(
        self,
        dim,
        max_position_embeddings=2048,
        base=10000,
        device=None,
        scaling_factor=1.0,
    ):
        self.scaling_factor = scaling_factor
        super().__init__(dim, max_position_embeddings, base, device)

    def _set_cos_sin_cache(self, seq_len, device, dtype):
        self.max_seq_len_cached = seq_len

        if seq_len > self.max_position_embeddings:
            base = self.base * (
                (self.scaling_factor * seq_len / self.max_position_embeddings) - (self.scaling_factor - 1)
            ) ** (self.dim / (self.dim - 2))
            inv_freq = 1.0 / (base ** (torch.arange(0, self.dim, 2).float().to(device) / self.dim))
            self.register_buffer("inv_freq", inv_freq, persistent=False)

        t = torch.arange(self.max_seq_len_cached, device=device, dtype=self.inv_freq.dtype)

        freqs = torch.outer(t, self.inv_freq)
        # Different from paper, but it uses a different permutation in order to obtain the same calculation
        emb = torch.cat((freqs, freqs), dim=-1)
        self.register_buffer("cos_cached", emb.cos().to(dtype), persistent=False)
        self.register_buffer("sin_cached", emb.sin().to(dtype), persistent=False)


# Inverse dim formula to find dim based on number of rotations
def yarn_find_correction_dim(num_rotations, dim, base=10000, max_position_embeddings=2048):
    return (dim * math.log(max_position_embeddings / (num_rotations * 2 * math.pi))) / (2 * math.log(base))


# Find dim range bounds based on rotations
def yarn_find_correction_range(low_rot, high_rot, dim, base=10000, max_position_embeddings=2048):
    low = math.floor(yarn_find_correction_dim(low_rot, dim, base, max_position_embeddings))
    high = math.ceil(yarn_find_correction_dim(high_rot, dim, base, max_position_embeddings))
    return max(low, 0), min(high, dim - 1)  # Clamp values just in case


def yarn_get_mscale(scale=1, mscale=1):
    if scale <= 1:
        return 1.0
    return 0.1 * mscale * math.log(scale) + 1.0


def yarn_linear_ramp_mask(min, max, dim):
    if min == max:
        max += 0.001  # Prevent singularity

    linear_func = (torch.arange(dim, dtype=torch.float32) - min) / (max - min)
    ramp_func = torch.clamp(linear_func, 0, 1)
    return ramp_func
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -



optimum/habana/transformers/models/deepseek_v3/modeling_deepseek_v3.py [216:270]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
    def __init__(
        self,
        dim,
        max_position_embeddings=2048,
        base=10000,
        device=None,
        scaling_factor=1.0,
    ):
        self.scaling_factor = scaling_factor
        super().__init__(dim, max_position_embeddings, base, device)

    def _set_cos_sin_cache(self, seq_len, device, dtype):
        self.max_seq_len_cached = seq_len

        if seq_len > self.max_position_embeddings:
            base = self.base * (
                (self.scaling_factor * seq_len / self.max_position_embeddings) - (self.scaling_factor - 1)
            ) ** (self.dim / (self.dim - 2))
            inv_freq = 1.0 / (base ** (torch.arange(0, self.dim, 2).float().to(device) / self.dim))
            self.register_buffer("inv_freq", inv_freq, persistent=False)

        t = torch.arange(self.max_seq_len_cached, device=device, dtype=self.inv_freq.dtype)

        freqs = torch.outer(t, self.inv_freq)
        # Different from paper, but it uses a different permutation in order to obtain the same calculation
        emb = torch.cat((freqs, freqs), dim=-1)
        self.register_buffer("cos_cached", emb.cos().to(dtype), persistent=False)
        self.register_buffer("sin_cached", emb.sin().to(dtype), persistent=False)


# Inverse dim formula to find dim based on number of rotations
def yarn_find_correction_dim(num_rotations, dim, base=10000, max_position_embeddings=2048):
    return (dim * math.log(max_position_embeddings / (num_rotations * 2 * math.pi))) / (2 * math.log(base))


# Find dim range bounds based on rotations
def yarn_find_correction_range(low_rot, high_rot, dim, base=10000, max_position_embeddings=2048):
    low = math.floor(yarn_find_correction_dim(low_rot, dim, base, max_position_embeddings))
    high = math.ceil(yarn_find_correction_dim(high_rot, dim, base, max_position_embeddings))
    return max(low, 0), min(high, dim - 1)  # Clamp values just in case


def yarn_get_mscale(scale=1, mscale=1):
    if scale <= 1:
        return 1.0
    return 0.1 * mscale * math.log(scale) + 1.0


def yarn_linear_ramp_mask(min, max, dim):
    if min == max:
        max += 0.001  # Prevent singularity

    linear_func = (torch.arange(dim, dtype=torch.float32) - min) / (max - min)
    ramp_func = torch.clamp(linear_func, 0, 1)
    return ramp_func
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -