diff options
| author | Jani Monoses <jani.monoses@gmail.com> | 2025-01-13 15:35:33 +0200 |
|---|---|---|
| committer | GitHub <noreply@github.com> | 2025-01-13 14:35:33 +0100 |
| commit | ab7ff7081eab36958b82b98b89cee3eacf877111 (patch) | |
| tree | a6768826a260a190bfe774fbbf954d6e85b1c5ae /candle-transformers | |
| parent | 461e8c1685e003bdddfd1e7d1aa5092786ca9df5 (diff) | |
| download | candle-ab7ff7081eab36958b82b98b89cee3eacf877111.tar.gz candle-ab7ff7081eab36958b82b98b89cee3eacf877111.tar.bz2 candle-ab7ff7081eab36958b82b98b89cee3eacf877111.zip | |
Fixes for running Phi-4 quantized. (#2714)
Diffstat (limited to 'candle-transformers')
| -rw-r--r-- | candle-transformers/src/models/quantized_phi3.rs | 2 |
1 files changed, 1 insertions, 1 deletions
diff --git a/candle-transformers/src/models/quantized_phi3.rs b/candle-transformers/src/models/quantized_phi3.rs index 51a75f38..1ceb48d1 100644 --- a/candle-transformers/src/models/quantized_phi3.rs +++ b/candle-transformers/src/models/quantized_phi3.rs @@ -127,7 +127,7 @@ impl LayerWeights { .reshape((b_sz, seq_len, self.n_head, self.head_dim))? .transpose(1, 2)?; let k = k - .reshape((b_sz, seq_len, self.n_head, self.head_dim))? + .reshape((b_sz, seq_len, self.n_kv_head, self.head_dim))? .transpose(1, 2)?; let v = v .reshape((b_sz, seq_len, self.n_kv_head, self.head_dim))? |