Skip to content

Commit

Permalink
update recipes for 9b and 27b
Browse files Browse the repository at this point in the history
  • Loading branch information
Optimox committed Oct 24, 2024
1 parent 6f89920 commit 0d53660
Show file tree
Hide file tree
Showing 8 changed files with 25 additions and 25 deletions.
6 changes: 3 additions & 3 deletions recipes/configs/gemma2/27B_full.yaml
Original file line number Diff line number Diff line change
Expand Up @@ -19,7 +19,7 @@
# Tokenizer
tokenizer:
_component_: torchtune.models.gemma.gemma_tokenizer
path: /tmp/gemma2-27b/tokenizer.model
path: /tmp/gemma-2-27b/tokenizer.model

# Dataset
dataset:
Expand All @@ -33,12 +33,12 @@ model:

checkpointer:
_component_: torchtune.training.FullModelHFCheckpointer
checkpoint_dir: /tmp/gemma2-27b/
checkpoint_dir: /tmp/gemma-2-27b/
checkpoint_files:
filename_format: model-{}-of-{}.safetensors
max_filename: 00024
recipe_checkpoint: null
output_dir: /tmp/gemma2-27b
output_dir: /tmp/gemma-2-27b
model_type: GEMMA2
resume_from_checkpoint: False

Expand Down
6 changes: 3 additions & 3 deletions recipes/configs/gemma2/27B_lora.yaml
Original file line number Diff line number Diff line change
Expand Up @@ -19,7 +19,7 @@
# Tokenizer
tokenizer:
_component_: torchtune.models.gemma.gemma_tokenizer
path: /tmp/gemma2-27b/tokenizer.model
path: /tmp/gemma-2-27b/tokenizer.model

# Dataset
dataset:
Expand All @@ -38,12 +38,12 @@ model:

checkpointer:
_component_: torchtune.training.FullModelHFCheckpointer
checkpoint_dir: /tmp/gemma2-27b/
checkpoint_dir: /tmp/gemma-2-27b/
checkpoint_files:
filename_format: model-{}-of-{}.safetensors
max_filename: 00024
recipe_checkpoint: null
output_dir: /tmp/gemma2-27b/
output_dir: /tmp/gemma-2-27b/
model_type: GEMMA2
resume_from_checkpoint: False
save_adapter_weights_only: False
Expand Down
6 changes: 3 additions & 3 deletions recipes/configs/gemma2/27B_lora_single_device.yaml
Original file line number Diff line number Diff line change
Expand Up @@ -18,7 +18,7 @@
# Tokenizer
tokenizer:
_component_: torchtune.models.gemma.gemma_tokenizer
path: /tmp/gemma2-27b/tokenizer.model
path: /tmp/gemma-2-27b/tokenizer.model

# Dataset
dataset:
Expand All @@ -37,12 +37,12 @@ model:

checkpointer:
_component_: torchtune.training.FullModelHFCheckpointer
checkpoint_dir: /tmp/gemma2-27b/
checkpoint_dir: /tmp/gemma-2-27b/
checkpoint_files:
filename_format: model-{}-of-{}.safetensors
max_filename: 00024
recipe_checkpoint: null
output_dir: /tmp/gemma2-27b/
output_dir: /tmp/gemma-2-27b/
model_type: GEMMA2
resume_from_checkpoint: False
save_adapter_weights_only: False
Expand Down
6 changes: 3 additions & 3 deletions recipes/configs/gemma2/27B_qlora_single_device.yaml
Original file line number Diff line number Diff line change
Expand Up @@ -18,7 +18,7 @@
# Tokenizer
tokenizer:
_component_: torchtune.models.gemma.gemma_tokenizer
path: /tmp/gemma2-27b/tokenizer.model
path: /tmp/gemma-2-27b/tokenizer.model

# Dataset
dataset:
Expand All @@ -37,12 +37,12 @@ model:

checkpointer:
_component_: torchtune.training.FullModelHFCheckpointer
checkpoint_dir: /tmp/gemma2-27b/
checkpoint_dir: /tmp/gemma-2-27b/
checkpoint_files:
filename_format: model-{}-of-{}.safetensors
max_filename: 00024
recipe_checkpoint: null
output_dir: /tmp/gemma2-27b/
output_dir: /tmp/gemma-2-27b/
model_type: GEMMA2
resume_from_checkpoint: False
save_adapter_weights_only: False
Expand Down
6 changes: 3 additions & 3 deletions recipes/configs/gemma2/9B_full.yaml
Original file line number Diff line number Diff line change
Expand Up @@ -19,7 +19,7 @@
# Tokenizer
tokenizer:
_component_: torchtune.models.gemma.gemma_tokenizer
path: /tmp/gemma2-9b/tokenizer.model
path: /tmp/gemma-2-9b/tokenizer.model

# Dataset
dataset:
Expand All @@ -33,12 +33,12 @@ model:

checkpointer:
_component_: torchtune.training.FullModelHFCheckpointer
checkpoint_dir: /tmp/gemma2-9b/
checkpoint_dir: /tmp/gemma-2-9b/
checkpoint_files:
filename_format: model-{}-of-{}.safetensors
max_filename: 00008
recipe_checkpoint: null
output_dir: /tmp/gemma2-9b
output_dir: /tmp/gemma-2-9b
model_type: GEMMA2
resume_from_checkpoint: False

Expand Down
6 changes: 3 additions & 3 deletions recipes/configs/gemma2/9B_lora.yaml
Original file line number Diff line number Diff line change
Expand Up @@ -19,7 +19,7 @@
# Tokenizer
tokenizer:
_component_: torchtune.models.gemma.gemma_tokenizer
path: /tmp/gemma2-9b/tokenizer.model
path: /tmp/gemma-2-9b/tokenizer.model

# Dataset
dataset:
Expand All @@ -38,12 +38,12 @@ model:

checkpointer:
_component_: torchtune.training.FullModelHFCheckpointer
checkpoint_dir: /tmp/gemma2-9b/
checkpoint_dir: /tmp/gemma-2-9b/
checkpoint_files:
filename_format: model-{}-of-{}.safetensors
max_filename: 00008
recipe_checkpoint: null
output_dir: /tmp/gemma2-9b/
output_dir: /tmp/gemma-2-9b/
model_type: GEMMA2
resume_from_checkpoint: False
save_adapter_weights_only: False
Expand Down
6 changes: 3 additions & 3 deletions recipes/configs/gemma2/9B_lora_single_device.yaml
Original file line number Diff line number Diff line change
Expand Up @@ -18,7 +18,7 @@
# Tokenizer
tokenizer:
_component_: torchtune.models.gemma.gemma_tokenizer
path: /tmp/gemma2-9b/tokenizer.model
path: /tmp/gemma-2-9b/tokenizer.model

# Dataset
dataset:
Expand All @@ -37,12 +37,12 @@ model:

checkpointer:
_component_: torchtune.training.FullModelHFCheckpointer
checkpoint_dir: /tmp/gemma2-9b/
checkpoint_dir: /tmp/gemma-2-9b/
checkpoint_files:
filename_format: model-{}-of-{}.safetensors
max_filename: 00008
recipe_checkpoint: null
output_dir: /tmp/gemma2-9b/
output_dir: /tmp/gemma-2-9b/
model_type: GEMMA2
resume_from_checkpoint: False
save_adapter_weights_only: False
Expand Down
8 changes: 4 additions & 4 deletions recipes/configs/gemma2/9B_qlora_single_device.yaml
Original file line number Diff line number Diff line change
Expand Up @@ -18,7 +18,7 @@
# Tokenizer
tokenizer:
_component_: torchtune.models.gemma.gemma_tokenizer
path: /tmp/gemma2-9b/tokenizer.model
path: /tmp/gemma-2-9b/tokenizer.model

# Dataset
dataset:
Expand All @@ -28,7 +28,7 @@ shuffle: True

# Model Arguments
model:
_component_: torchtune.models.gemma2.qlora_gemma_9b
_component_: torchtune.models.gemma2.qlora_gemma2_9b
lora_attn_modules: ['q_proj', 'k_proj', 'v_proj']
apply_lora_to_mlp: True
lora_rank: 64
Expand All @@ -37,12 +37,12 @@ model:

checkpointer:
_component_: torchtune.training.FullModelHFCheckpointer
checkpoint_dir: /tmp/gemma2-9b/
checkpoint_dir: /tmp/gemma-2-9b/
checkpoint_files:
filename_format: model-{}-of-{}.safetensors
max_filename: 00008
recipe_checkpoint: null
output_dir: /tmp/gemma2-9b/
output_dir: /tmp/gemma-2-9b/
model_type: GEMMA2
resume_from_checkpoint: False
save_adapter_weights_only: False
Expand Down

0 comments on commit 0d53660

Please sign in to comment.