Add LM head quantization support for ModelOpt (#42124)
Signed-off-by: weimingc <17592131+meenchen@users.noreply.github.com>
This commit is contained in:
@@ -875,6 +875,7 @@ class NemotronHForCausalLM(
|
||||
self.lm_head = ParallelLMHead(
|
||||
config.vocab_size,
|
||||
config.hidden_size,
|
||||
quant_config=self.quant_config,
|
||||
prefix=maybe_prefix(prefix, "lm_head"),
|
||||
)
|
||||
|
||||
|
||||
@@ -477,6 +477,7 @@ class Qwen3_5ForCausalLMBase(
|
||||
self.lm_head = ParallelLMHead(
|
||||
config.vocab_size,
|
||||
config.hidden_size,
|
||||
quant_config=self.quant_config,
|
||||
prefix=maybe_prefix(prefix, "lm_head"),
|
||||
)
|
||||
else:
|
||||
|
||||
@@ -381,6 +381,7 @@ class Qwen3_5MTP(nn.Module, SupportsMultiModal):
|
||||
self.lm_head = ParallelLMHead(
|
||||
config.vocab_size,
|
||||
config.hidden_size,
|
||||
quant_config=self.quant_config,
|
||||
prefix=maybe_prefix(prefix, "lm_head"),
|
||||
)
|
||||
else:
|
||||
|
||||
Reference in New Issue
Block a user