nikola/llama_cpp
1
0
Fork 0
Code Issues Pull Requests Actions 11 Packages Projects Releases Wiki Activity

dflash: refactor draft model conversion (#25110)

Browse Source 
* dflash: refactor draft model conversion

* apply fix for eagle3 convert
This commit is contained in:
Ruixiang Wang
2026-06-28 20:31:48 +02:00
committed by GitHub
parent c818263f2a
commit fa72bc6826
5 changed files with 31 additions and 17 deletions
Download Patch File Download Diff File Expand all files Collapse all files
conversion/llama.py
+3 -3
View File
@@ -73,7 +73,7 @@ class LlamaModel(TextModel):
target_num_layers = target_config["num_hidden_layers"]
target_layers = [2, target_num_layers // 2, target_num_layers - 3]
logger.info(f"EAGLE-3: target_layers = {target_layers} (target model has {target_num_layers} layers)")
self.gguf_writer.add_array(f"{self.gguf_writer.arch}.target_layers", target_layers)
self.gguf_writer.add_target_layers(target_layers)
# target_hidden_size: prefer eagle3 config, fallback to target config
if eagle3_raw_config.get("target_hidden_size") is not None:
@@ -83,12 +83,12 @@ class LlamaModel(TextModel):
target_hidden_size = target_config["hidden_size"]
src = "target model config"
logger.info(f"EAGLE-3: target_hidden_size = {target_hidden_size} (from {src})")
self.gguf_writer.add_uint32(f"{self.gguf_writer.arch}.target_hidden_size", target_hidden_size)
self.gguf_writer.add_target_hidden_size(target_hidden_size)
# norm_before_residual (RedHat-style eagle3 specific)
norm_before_residual = eagle3_raw_config.get("norm_before_residual", False)
logger.info(f"EAGLE-3: norm_before_residual = {norm_before_residual}")
self.gguf_writer.add_bool(f"{self.gguf_writer.arch}.norm_before_residual", norm_before_residual)
self.gguf_writer.add_norm_before_residual(norm_before_residual)
def set_vocab(self):
# eagle3: use tokenizer from target model if provided