Upload folder using huggingface_hub
Browse files- .gitattributes +3 -0
- ELF-B-de-en.yml +70 -0
- checkpoint_0/_CHECKPOINT_METADATA +1 -0
- checkpoint_0/_METADATA +0 -0
- checkpoint_0/d/c602cf9787ff97fc0f5bb3081af834b5 +0 -0
- checkpoint_0/manifest.ocdbt +0 -0
- checkpoint_0/ocdbt.process_0/d/1b4dc63bea2d98dc392e69e33f2ebe7a +3 -0
- checkpoint_0/ocdbt.process_0/d/ad715c4692f9b6da91a21655a0bb3af0 +3 -0
- checkpoint_0/ocdbt.process_0/d/b7360d60d79106040c95d6b7abb74ef6 +0 -0
- checkpoint_0/ocdbt.process_0/d/dbc8b9f046ec6f3845cf6c9d8c255b93 +3 -0
- checkpoint_0/ocdbt.process_0/manifest.ocdbt +0 -0
.gitattributes
CHANGED
|
@@ -33,3 +33,6 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
|
|
| 33 |
*.zip filter=lfs diff=lfs merge=lfs -text
|
| 34 |
*.zst filter=lfs diff=lfs merge=lfs -text
|
| 35 |
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
|
|
|
|
|
|
|
|
|
|
|
| 33 |
*.zip filter=lfs diff=lfs merge=lfs -text
|
| 34 |
*.zst filter=lfs diff=lfs merge=lfs -text
|
| 35 |
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
| 36 |
+
checkpoint_0/ocdbt.process_0/d/1b4dc63bea2d98dc392e69e33f2ebe7a filter=lfs diff=lfs merge=lfs -text
|
| 37 |
+
checkpoint_0/ocdbt.process_0/d/ad715c4692f9b6da91a21655a0bb3af0 filter=lfs diff=lfs merge=lfs -text
|
| 38 |
+
checkpoint_0/ocdbt.process_0/d/dbc8b9f046ec6f3845cf6c9d8c255b93 filter=lfs diff=lfs merge=lfs -text
|
ELF-B-de-en.yml
ADDED
|
@@ -0,0 +1,70 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# Dataset
|
| 2 |
+
data_path: "embedded-language-flows/wmt14_de-en_train_t5"
|
| 3 |
+
eval_data_path: "embedded-language-flows/wmt14_de-en_validation_t5"
|
| 4 |
+
max_length: 128
|
| 5 |
+
max_input_length: 64
|
| 6 |
+
pad_token: eos
|
| 7 |
+
|
| 8 |
+
# Encoder
|
| 9 |
+
encoder_model_name: t5-small
|
| 10 |
+
encoder_checkpoint: "embedded-language-flows/t5_small_encoder_jax/t5_small_encoder_jax.pkl"
|
| 11 |
+
latent_mean: 0.0
|
| 12 |
+
latent_std: 0.2
|
| 13 |
+
|
| 14 |
+
# Model architecture
|
| 15 |
+
model: ELF-B # ELF-B, ELF-M
|
| 16 |
+
bottleneck_dim: 128
|
| 17 |
+
num_time_tokens: 4
|
| 18 |
+
num_self_cond_cfg_tokens: 4
|
| 19 |
+
num_model_mode_tokens: 4
|
| 20 |
+
|
| 21 |
+
# Denoiser objective
|
| 22 |
+
denoiser_p_mean: -1.5
|
| 23 |
+
denoiser_p_std: 0.8
|
| 24 |
+
denoiser_noise_scale: 2.0
|
| 25 |
+
t_eps: 0.05
|
| 26 |
+
time_schedule: "logit_normal"
|
| 27 |
+
|
| 28 |
+
# Decoder objective
|
| 29 |
+
decoder_prob: 0.2
|
| 30 |
+
decoder_noise_scale: 5.0
|
| 31 |
+
decoder_p_mean: 0.8
|
| 32 |
+
decoder_p_std: 0.8
|
| 33 |
+
|
| 34 |
+
# Conditioning / CFG
|
| 35 |
+
label_drop_prob: 0.1
|
| 36 |
+
self_cond_prob: 0.5
|
| 37 |
+
|
| 38 |
+
# Training (optimizer + schedule)
|
| 39 |
+
epochs: 100
|
| 40 |
+
global_batch_size: 512
|
| 41 |
+
blr: 0.001
|
| 42 |
+
weight_decay: 0.0
|
| 43 |
+
warmup_steps: 5000
|
| 44 |
+
optimzier: muon
|
| 45 |
+
|
| 46 |
+
# EMA
|
| 47 |
+
ema_decay1: 0.9999
|
| 48 |
+
|
| 49 |
+
# Sampling
|
| 50 |
+
sampling_configs_path: "configs/sampling_configs/cond_sampling_configs.yml"
|
| 51 |
+
num_samples: 5000
|
| 52 |
+
|
| 53 |
+
# Logging & Checkpointing
|
| 54 |
+
log_freq: 100
|
| 55 |
+
save_freq: 10
|
| 56 |
+
save_last_freq: 1000
|
| 57 |
+
eval_freq: 10
|
| 58 |
+
|
| 59 |
+
# Output
|
| 60 |
+
output_dir: "outputs/elf_b-wmt-de-en"
|
| 61 |
+
resume: null
|
| 62 |
+
|
| 63 |
+
# Wandb
|
| 64 |
+
use_wandb: true
|
| 65 |
+
wandb_project: elf
|
| 66 |
+
wandb_entity:
|
| 67 |
+
wandb_run_name: elf_b-wmt-de-en
|
| 68 |
+
|
| 69 |
+
# Misc
|
| 70 |
+
seed: 42
|
checkpoint_0/_CHECKPOINT_METADATA
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
{"init_timestamp_nsecs": 1778515893369740925, "commit_timestamp_nsecs": 1778515894613494062}
|
checkpoint_0/_METADATA
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
checkpoint_0/d/c602cf9787ff97fc0f5bb3081af834b5
ADDED
|
Binary file (14.4 kB). View file
|
|
|
checkpoint_0/manifest.ocdbt
ADDED
|
Binary file (118 Bytes). View file
|
|
|
checkpoint_0/ocdbt.process_0/d/1b4dc63bea2d98dc392e69e33f2ebe7a
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:047893f95a836893ea82af14022afe0eb02ebdbe1eb5adaac2449fbbfe2a5ac8
|
| 3 |
+
size 423174144
|
checkpoint_0/ocdbt.process_0/d/ad715c4692f9b6da91a21655a0bb3af0
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b118fc32474de8b0c152fd7ac4c459efd04f7e611151187b7e4c354a4c18aff6
|
| 3 |
+
size 21368832
|
checkpoint_0/ocdbt.process_0/d/b7360d60d79106040c95d6b7abb74ef6
ADDED
|
Binary file (178 Bytes). View file
|
|
|
checkpoint_0/ocdbt.process_0/d/dbc8b9f046ec6f3845cf6c9d8c255b93
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:daf707007b8e2b42aaa791869876b4f21fcf562ef6e3812db37326f8e4fb0cf2
|
| 3 |
+
size 332054528
|
checkpoint_0/ocdbt.process_0/manifest.ocdbt
ADDED
|
Binary file (278 Bytes). View file
|
|
|