remove apply lora to output from config

pytorch · Nov 7, 2024 · 068ed6f · 068ed6f
1 parent 1c1e0f4
commit 068ed6f
Show file tree

Hide file tree

Showing 12 changed files with 0 additions and 12 deletions.
diff --git a/recipes/configs/llama3_2/1B_lora.yaml b/recipes/configs/llama3_2/1B_lora.yaml
@@ -28,7 +28,6 @@ model:
   _component_: torchtune.models.llama3_2.lora_llama3_2_1b
   lora_attn_modules: ['q_proj', 'v_proj', 'output_proj']
   apply_lora_to_mlp: True
-  apply_lora_to_output: False
   lora_rank: 64  # higher increases accuracy and memory
   lora_alpha: 128  # usually alpha=2*rank
   lora_dropout: 0.0

diff --git a/recipes/configs/llama3_2/1B_lora_single_device.yaml b/recipes/configs/llama3_2/1B_lora_single_device.yaml
@@ -21,7 +21,6 @@ model:
   _component_: torchtune.models.llama3_2.lora_llama3_2_1b
   lora_attn_modules: ['q_proj', 'v_proj', 'output_proj']
   apply_lora_to_mlp: True
-  apply_lora_to_output: False
   lora_rank: 64  # higher increases accuracy and memory
   lora_alpha: 128  # usually alpha=2*rank
   lora_dropout: 0.0

diff --git a/recipes/configs/llama3_2/1B_qlora_single_device.yaml b/recipes/configs/llama3_2/1B_qlora_single_device.yaml
@@ -20,7 +20,6 @@ model:
   _component_: torchtune.models.llama3_2.qlora_llama3_2_1b
   lora_attn_modules: ['q_proj', 'v_proj', 'output_proj']
   apply_lora_to_mlp: True
-  apply_lora_to_output: False
   lora_rank: 64  # higher increases accuracy and memory
   lora_alpha: 128  # usually alpha=2*rank
   lora_dropout: 0.0

diff --git a/recipes/configs/llama3_2/3B_lora.yaml b/recipes/configs/llama3_2/3B_lora.yaml
@@ -28,7 +28,6 @@ model:
   _component_: torchtune.models.llama3_2.lora_llama3_2_3b
   lora_attn_modules: ['q_proj', 'v_proj', 'output_proj']
   apply_lora_to_mlp: True
-  apply_lora_to_output: False
   lora_rank: 64  # higher increases accuracy and memory
   lora_alpha: 128  # usually alpha=2*rank
   lora_dropout: 0.0

diff --git a/recipes/configs/llama3_2/3B_lora_single_device.yaml b/recipes/configs/llama3_2/3B_lora_single_device.yaml
@@ -21,7 +21,6 @@ model:
   _component_: torchtune.models.llama3_2.lora_llama3_2_3b
   lora_attn_modules: ['q_proj', 'v_proj', 'output_proj']
   apply_lora_to_mlp: True
-  apply_lora_to_output: False
   lora_rank: 64  # higher increases accuracy and memory
   lora_alpha: 128  # usually alpha=2*rank
   lora_dropout: 0.0

diff --git a/recipes/configs/llama3_2/3B_qlora_single_device.yaml b/recipes/configs/llama3_2/3B_qlora_single_device.yaml
@@ -20,7 +20,6 @@ model:
   _component_: torchtune.models.llama3_2.qlora_llama3_2_3b
   lora_attn_modules: ['q_proj', 'v_proj', 'output_proj']
   apply_lora_to_mlp: True
-  apply_lora_to_output: False
   lora_rank: 64  # higher increases accuracy and memory
   lora_alpha: 128  # usually alpha=2*rank
   lora_dropout: 0.0

diff --git a/recipes/configs/qwen2_5/0_5B_lora.yaml b/recipes/configs/qwen2_5/0_5B_lora.yaml
@@ -22,7 +22,6 @@ model:
   _component_: torchtune.models.qwen2_5.lora_qwen2_5_0_5b
   lora_attn_modules: ['q_proj', 'v_proj', 'output_proj']
   apply_lora_to_mlp: True
-  apply_lora_to_output: False
   lora_rank: 32  # higher increases accuracy and memory
   lora_alpha: 64  # usually alpha=2*rank
   lora_dropout: 0.0

diff --git a/recipes/configs/qwen2_5/0_5B_lora_single_device.yaml b/recipes/configs/qwen2_5/0_5B_lora_single_device.yaml
@@ -21,7 +21,6 @@ model:
   _component_: torchtune.models.qwen2_5.lora_qwen2_5_0_5b
   lora_attn_modules: ['q_proj', 'v_proj', 'output_proj']
   apply_lora_to_mlp: True
-  apply_lora_to_output: False
   lora_rank: 32  # higher increases accuracy and memory
   lora_alpha: 64  # usually alpha=2*rank
   lora_dropout: 0.0

diff --git a/recipes/configs/qwen2_5/1_5B_lora.yaml b/recipes/configs/qwen2_5/1_5B_lora.yaml
@@ -22,7 +22,6 @@ model:
   _component_: torchtune.models.qwen2_5.lora_qwen2_5_1_5b_instruct
   lora_attn_modules: ['q_proj', 'v_proj', 'output_proj']
   apply_lora_to_mlp: True
-  apply_lora_to_output: False
   lora_rank: 32  # higher increases accuracy and memory
   lora_alpha: 64  # usually alpha=2*rank
   lora_dropout: 0.0

diff --git a/recipes/configs/qwen2_5/1_5B_lora_single_device.yaml b/recipes/configs/qwen2_5/1_5B_lora_single_device.yaml
@@ -21,7 +21,6 @@ model:
   _component_: torchtune.models.qwen2_5.lora_qwen2_5_1_5b_instruct
   lora_attn_modules: ['q_proj', 'v_proj', 'output_proj']
   apply_lora_to_mlp: True
-  apply_lora_to_output: False
   lora_rank: 32  # higher increases accuracy and memory
   lora_alpha: 64  # usually alpha=2*rank
   lora_dropout: 0.0

diff --git a/recipes/configs/qwen2_5/3B_lora.yaml b/recipes/configs/qwen2_5/3B_lora.yaml
@@ -22,7 +22,6 @@ model:
   _component_: torchtune.models.qwen2_5.lora_qwen2_5_3b
   lora_attn_modules: ['q_proj', 'v_proj', 'output_proj']
   apply_lora_to_mlp: True
-  apply_lora_to_output: False
   lora_rank: 8  # higher increases accuracy and memory
   lora_alpha: 16  # usually alpha=2*rank
   lora_dropout: 0.0

diff --git a/recipes/configs/qwen2_5/3B_lora_single_device.yaml b/recipes/configs/qwen2_5/3B_lora_single_device.yaml
@@ -21,7 +21,6 @@ model:
   _component_: torchtune.models.qwen2_5.lora_qwen2_5_3b
   lora_attn_modules: ['q_proj', 'v_proj', 'output_proj']
   apply_lora_to_mlp: True
-  apply_lora_to_output: False
   lora_rank: 8  # higher increases accuracy and memory
   lora_alpha: 16  # usually alpha=2*rank
   lora_dropout: 0.0