apunkt/doc-to-lora
1
0
Fork 0
mirror of https://github.com/SakanaAI/doc-to-lora.git synced 2026-07-26 17:11:02 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions

add repeat prompt

Browse source
This commit is contained in:
51616 2025-01-08 16:17:07 +00:00
parent 1699668aef
commit 2381b6a93e
17 changed files with 37 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

1
configs/context_numbers_10.yaml
View file

@ -2,6 +2,7 @@ output_dir: "" # just a placeholder
bf16: true
model_name_or_path: meta-llama/Llama-3.2-1B-Instruct
label_names: ["labels"]
add_repeat_prompt: false
# eval_on_start: True
# eval_strategy: "steps"
# eval_steps: 500

1
configs/context_numbers_128.yaml
View file

@ -2,6 +2,7 @@ output_dir: "" # just a placeholder
bf16: true
model_name_or_path: meta-llama/Llama-3.2-1B-Instruct
label_names: ["labels"]
add_repeat_prompt: false
# eval_on_start: True
# eval_strategy: "steps"
# eval_steps: 500

1
configs/context_numbers_128_only.yaml
View file

@ -2,6 +2,7 @@ output_dir: "" # just a placeholder
bf16: true
model_name_or_path: meta-llama/Llama-3.2-1B-Instruct
label_names: ["labels"]
add_repeat_prompt: false
# eval_on_start: True
# eval_strategy: "steps"
# eval_steps: 500

1
configs/context_numbers_256.yaml
View file

@ -2,6 +2,7 @@ output_dir: "" # just a placeholder
bf16: true
model_name_or_path: meta-llama/Llama-3.2-1B-Instruct
label_names: ["labels"]
add_repeat_prompt: false
# eval_on_start: True
# eval_strategy: "steps"
# eval_steps: 500

1
configs/context_numbers_256_only.yaml
View file

@ -2,6 +2,7 @@ output_dir: "" # just a placeholder
bf16: true
model_name_or_path: meta-llama/Llama-3.2-1B-Instruct
label_names: ["labels"]
add_repeat_prompt: false
# eval_on_start: True
# eval_strategy: "steps"
# eval_steps: 500

1
configs/context_numbers_32.yaml
View file

@ -2,6 +2,7 @@ output_dir: "" # just a placeholder
bf16: true
model_name_or_path: meta-llama/Llama-3.2-1B-Instruct
label_names: ["labels"]
add_repeat_prompt: false
# eval_on_start: True
# eval_strategy: "steps"
# eval_steps: 500

1
configs/context_numbers_32_only.yaml
View file

@ -2,6 +2,7 @@ output_dir: "" # just a placeholder
bf16: true
model_name_or_path: meta-llama/Llama-3.2-1B-Instruct
label_names: ["labels"]
add_repeat_prompt: false
# eval_on_start: True
# eval_strategy: "steps"
# eval_steps: 500

1
configs/context_numbers_64.yaml
View file

@ -2,6 +2,7 @@ output_dir: "" # just a placeholder
bf16: true
model_name_or_path: meta-llama/Llama-3.2-1B-Instruct
label_names: ["labels"]
add_repeat_prompt: false
# eval_on_start: True
# eval_strategy: "steps"
# eval_steps: 500

1
configs/context_numbers_64_only.yaml
View file

@ -2,6 +2,7 @@ output_dir: "" # just a placeholder
bf16: true
model_name_or_path: meta-llama/Llama-3.2-1B-Instruct
label_names: ["labels"]
add_repeat_prompt: false
# eval_on_start: True
# eval_strategy: "steps"
# eval_steps: 500

1
configs/context_numbers_debug.yaml
View file

@ -2,6 +2,7 @@ output_dir: "" # just a placeholder
bf16: true
model_name_or_path: meta-llama/Llama-3.2-1B-Instruct
label_names: ["labels"]
add_repeat_prompt: false
eval_on_start: True
eval_strategy: "steps"
eval_steps: 500

1
configs/context_numbers_easy.yaml
View file

@ -2,6 +2,7 @@ output_dir: "" # just a placeholder
bf16: true
model_name_or_path: meta-llama/Llama-3.2-1B-Instruct
label_names: ["labels"]
add_repeat_prompt: false
eval_on_start: True
eval_strategy: "steps"
eval_steps: 500

1
configs/default.yaml
View file

@ -2,6 +2,7 @@ output_dir: "" # just a placeholder
bf16: true
model_name_or_path: meta-llama/Llama-3.2-1B-Instruct
label_names: ["labels"]
add_repeat_prompt: false
eval_on_start: True
eval_strategy: "steps"
eval_steps: 500

2
configs/pwc.yaml
View file

@ -18,7 +18,7 @@ label_names: ["labels"]
per_device_train_batch_size: 32
per_device_eval_batch_size: 32
max_val_samples_per_ds: 20
max_val_samples_per_ds: 1000
# optim: schedule_free_adamw
learning_rate: 0.00001
# lr_scheduler_type: "constant_with_warmup"

3
configs/pwc_and_ctx_numbers_256.yaml
View file

@ -2,6 +2,7 @@ output_dir: "" # just a placeholder
bf16: true
model_name_or_path: meta-llama/Llama-3.2-1B-Instruct
label_names: ["labels"]
add_repeat_prompt: false
# eval_on_start: True
# eval_strategy: "steps"
# eval_steps: 500
@ -18,7 +19,7 @@ label_names: ["labels"]
per_device_train_batch_size: 32
per_device_eval_batch_size: 1
max_val_samples_per_ds: 20
max_val_samples_per_ds: 1000
# optim: schedule_free_adamw
learning_rate: 0.00001
# lr_scheduler_type: "constant_with_warmup"