ThatAIguyA world of ideas

THE CONSCIOUSNESS ATLAS
Read. Experiment. Follow the evidence.

ConsciousnessLearn · Explore · Joseph Anady

DeepSeek RotaryEmbedding._set_cos_sin_cache: t

In DeepseekV3RotaryEmbedding._set_cos_sin_cache, this statement defines or updates t. This is part of the positional-phase calculation. Frequency, position, sine/cosine and coordinate ordering must agree with the source rotary convention.

Donor model implementationhistorical source recovered; expression not executed in this website audit
Learn with the Atlas

Learn to read this model

Start with the symbols, follow the operations, and check what the result means.

Open guided lesson
donor-eq-f88d004c68957asource-code mathematical expression
t=torch.arange(self.maxseqlencached,device=device,dtype=self.invfreq.dtype)t = torch.arange(self.max_seq_len_cached, device=device, dtype=self.inv_freq.dtype)

Read the equation

In DeepseekV3RotaryEmbedding._set_cos_sin_cache, this statement defines or updates t. This is part of the positional-phase calculation. Frequency, position, sine/cosine and coordinate ordering must agree with the source rotary convention.

Terms & definitions

The recovered passage does not supply a separate symbol table. Read the source context for the notation used in this version; familiar letters may carry different meanings in other builds.

How it works

  1. Read t as the quantity being defined or updated by this relation.

  2. Substitute the source-defined variables or recorded values and follow the displayed arithmetic and indexing exactly.

  3. Use this version’s variable definitions and update order. The same symbol may be assigned a different meaning in another build.

Assumptions & limits

Source recovery establishes that the expression appears in this record. Its mathematical validity, implementation behavior, and interpretation must be assessed separately. See the linked source and research notes for the checks available for this version.

Original context

Read the surrounding source passage
def _set_cos_sin_cache(self, seq_len, device, dtype):
        self.max_seq_len_cached = seq_len
        t = torch.arange(self.max_seq_len_cached,
                         device=device,
                         dtype=self.inv_freq.dtype)

        freqs = torch.outer(t, self.inv_freq.to(t.device))
        # Different from paper, but it uses a different permutation in order to obtain the same calculation
Exact original notation
t = torch.arange(self.max_seq_len_cached,
                         device=device,
                         dtype=self.inv_freq.dtype)

Related builds

Sources & provenance

01
kimi source tree.txt
lines 2441–2443 · Aug 5, 2026
t = torch.arange(self.max_seq_len_cached, device=device, dtype=self.inv_freq.dtype)
02
kimi source tree.txt
lines 2485–2487 · Aug 5, 2026
t = torch.arange(self.max_seq_len_cached, device=device, dtype=self.inv_freq.dtype)
03
kimi source tree.txt
lines 2528–2530 · Aug 5, 2026
t = torch.arange(self.max_seq_len_cached, device=device, dtype=self.inv_freq.dtype)
04
DEEPSEEK_R1_COMPLETE.txt
lines 2246–2248 · Aug 5, 2026
t = torch.arange( self.max_seq_len_cached, device=device, dtype=self.inv_freq.dtype )
05
DEEPSEEK_R1_COMPLETE.txt
lines 2284–2286 · Aug 5, 2026
t = torch.arange( self.max_seq_len_cached, device=device, dtype=self.inv_freq.dtype )
06
DEEPSEEK_R1_COMPLETE.txt
lines 2324–2326 · Aug 5, 2026
t = torch.arange( self.max_seq_len_cached, device=device, dtype=self.inv_freq.dtype )

Search equations, builds, manuscript text, and recovered sources.