add a new version called V3_TIKTOKEN.

other edits based on suggestions.
apple · Nov 20, 2024 · f1362fa · f1362fa
1 parent b4d139e
commit f1362fa
Show file tree

Hide file tree

Showing 64 changed files with 87 additions and 3,131 deletions.
diff --git a/...fuji-1B-v3-flash-tiktoken-single-host.txt → ...fuji-1B-v3-tiktoken-flash-single-host.txt b/...fuji-1B-v3-flash-tiktoken-single-host.txt → ...fuji-1B-v3-tiktoken-flash-single-host.txt
diff --git a/...1B-v3-flash-tiktoken-single-host_init.txt → ...1B-v3-tiktoken-flash-single-host_init.txt b/...1B-v3-flash-tiktoken-single-host_init.txt → ...1B-v3-tiktoken-flash-single-host_init.txt
diff --git a/...lash-tiktoken-single-host_regularizer.txt → ...iktoken-flash-single-host_regularizer.txt b/...lash-tiktoken-single-host_regularizer.txt → ...iktoken-flash-single-host_regularizer.txt
diff --git a/....c4_trainer/fuji-1B-v3-flash-tiktoken.txt → ....c4_trainer/fuji-1B-v3-tiktoken-flash.txt b/....c4_trainer/fuji-1B-v3-flash-tiktoken.txt → ....c4_trainer/fuji-1B-v3-tiktoken-flash.txt
diff --git a/...rainer/fuji-1B-v3-flash-tiktoken_init.txt → ...rainer/fuji-1B-v3-tiktoken-flash_init.txt b/...rainer/fuji-1B-v3-flash-tiktoken_init.txt → ...rainer/fuji-1B-v3-tiktoken-flash_init.txt
diff --git a/...fuji-1B-v3-flash-tiktoken_regularizer.txt → ...fuji-1B-v3-tiktoken-flash_regularizer.txt b/...fuji-1B-v3-flash-tiktoken_regularizer.txt → ...fuji-1B-v3-tiktoken-flash_regularizer.txt
diff --git a/...fuji-3B-v3-flash-tiktoken-single-host.txt → ...fuji-3B-v3-tiktoken-flash-single-host.txt b/...fuji-3B-v3-flash-tiktoken-single-host.txt → ...fuji-3B-v3-tiktoken-flash-single-host.txt
diff --git a/...3B-v3-flash-tiktoken-single-host_init.txt → ...3B-v3-tiktoken-flash-single-host_init.txt b/...3B-v3-flash-tiktoken-single-host_init.txt → ...3B-v3-tiktoken-flash-single-host_init.txt
diff --git a/...lash-tiktoken-single-host_regularizer.txt → ...iktoken-flash-single-host_regularizer.txt b/...lash-tiktoken-single-host_regularizer.txt → ...iktoken-flash-single-host_regularizer.txt
diff --git a/....c4_trainer/fuji-3B-v3-flash-tiktoken.txt → ....c4_trainer/fuji-3B-v3-tiktoken-flash.txt b/....c4_trainer/fuji-3B-v3-flash-tiktoken.txt → ....c4_trainer/fuji-3B-v3-tiktoken-flash.txt
diff --git a/...rainer/fuji-3B-v3-flash-tiktoken_init.txt → ...rainer/fuji-3B-v3-tiktoken-flash_init.txt b/...rainer/fuji-3B-v3-flash-tiktoken_init.txt → ...rainer/fuji-3B-v3-tiktoken-flash_init.txt
diff --git a/...fuji-3B-v3-flash-tiktoken_regularizer.txt → ...fuji-3B-v3-tiktoken-flash_regularizer.txt b/...fuji-3B-v3-flash-tiktoken_regularizer.txt → ...fuji-3B-v3-tiktoken-flash_regularizer.txt
diff --git a/...iments/testdata/axlearn.experiments.text.gpt.c4_trainer/fuji-70B-v1-flash-single-host.txt b/...iments/testdata/axlearn.experiments.text.gpt.c4_trainer/fuji-70B-v1-flash-single-host.txt
diff --git a/...s/testdata/axlearn.experiments.text.gpt.c4_trainer/fuji-70B-v1-flash-single-host_init.txt b/...s/testdata/axlearn.experiments.text.gpt.c4_trainer/fuji-70B-v1-flash-single-host_init.txt
diff --git a/.../experiments/testdata/axlearn.experiments.text.gpt.c4_trainer/fuji-70B-v1-single-host.txt b/.../experiments/testdata/axlearn.experiments.text.gpt.c4_trainer/fuji-70B-v1-single-host.txt
diff --git a/...riments/testdata/axlearn.experiments.text.gpt.c4_trainer/fuji-70B-v1-single-host_init.txt b/...riments/testdata/axlearn.experiments.text.gpt.c4_trainer/fuji-70B-v1-single-host_init.txt
diff --git a/...iments/testdata/axlearn.experiments.text.gpt.c4_trainer/fuji-70B-v2-flash-single-host.txt b/...iments/testdata/axlearn.experiments.text.gpt.c4_trainer/fuji-70B-v2-flash-single-host.txt
diff --git a/...s/testdata/axlearn.experiments.text.gpt.c4_trainer/fuji-70B-v2-flash-single-host_init.txt b/...s/testdata/axlearn.experiments.text.gpt.c4_trainer/fuji-70B-v2-flash-single-host_init.txt
diff --git a/.../experiments/testdata/axlearn.experiments.text.gpt.c4_trainer/fuji-70B-v2-single-host.txt b/.../experiments/testdata/axlearn.experiments.text.gpt.c4_trainer/fuji-70B-v2-single-host.txt
diff --git a/...riments/testdata/axlearn.experiments.text.gpt.c4_trainer/fuji-70B-v2-single-host_init.txt b/...riments/testdata/axlearn.experiments.text.gpt.c4_trainer/fuji-70B-v2-single-host_init.txt