rubentium commited on
Commit
78bad5a
·
verified ·
1 Parent(s): 5cc0d45

Upload Part 5 Dense16k seed 42 final checkpoint step 22000

Browse files
checkpoints/FINAL4_20260927/part_5/365m_dense16k_qkgainfp32_s42/checkpoint-22000-tamia/config.json ADDED
@@ -0,0 +1,42 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "vocab_size": 151936,
3
+ "hidden_size": 768,
4
+ "intermediate_size": 3072,
5
+ "num_hidden_layers": 28,
6
+ "num_attention_heads": 6,
7
+ "num_key_value_heads": 2,
8
+ "head_dim": 128,
9
+ "hidden_act": "silu",
10
+ "max_position_embeddings": 32768,
11
+ "initializer_range": 0.02,
12
+ "rms_norm_eps": 1e-06,
13
+ "use_cache": true,
14
+ "tie_word_embeddings": true,
15
+ "rope_theta": 1000000.0,
16
+ "rope_scaling": null,
17
+ "positional": "rope",
18
+ "qk_norm": "on",
19
+ "qk_gain_fp32": true,
20
+ "norm_gain_fp32": false,
21
+ "qk_gain_mode": "vector",
22
+ "rope_train_position_scale_set": null,
23
+ "attention_bias": false,
24
+ "attention_dropout": 0.0,
25
+ "bos_token_id": 151643,
26
+ "eos_token_id": 151643,
27
+ "checkpoint_activations": false,
28
+ "attention_type": "dense",
29
+ "sparse_block_size": 128,
30
+ "sparse_num_random_blocks": 1,
31
+ "sparse_mask_pool_size": 1000,
32
+ "sparse_max_seq_length": 16384,
33
+ "sparse_layer_indices": null,
34
+ "sparse_mask_cache_dir": null,
35
+ "sparse_mask_seed": 0,
36
+ "sparse_mixed_nrb_set": null,
37
+ "moba_block_size": 128,
38
+ "moba_top_k": 32,
39
+ "asentmax_alpha": 1.5,
40
+ "asentmax_learn_temp": true,
41
+ "asentmax_init_temp": 1.0
42
+ }
checkpoints/FINAL4_20260927/part_5/365m_dense16k_qkgainfp32_s42/checkpoint-22000-tamia/loader_gen_state_rank0.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3f59a1416dd062038a7e2dbac8c330163f45326660faeb92a199f62460447539
3
+ size 6802
checkpoints/FINAL4_20260927/part_5/365m_dense16k_qkgainfp32_s42/checkpoint-22000-tamia/loader_gen_state_rank1.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9907050ee7aa7455a32aae231218aac0666dd0bc7b78294eeb2cb3adc77e63c8
3
+ size 6802
checkpoints/FINAL4_20260927/part_5/365m_dense16k_qkgainfp32_s42/checkpoint-22000-tamia/loader_gen_state_rank2.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3adc40f4c25919716a49dee06d626601e0abca351145ddf1ed45e0bbfca3ea0c
3
+ size 6802
checkpoints/FINAL4_20260927/part_5/365m_dense16k_qkgainfp32_s42/checkpoint-22000-tamia/loader_gen_state_rank3.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5e32dfbaa9aad434214e9087e9315fae33fa445ebde26c320765a117f68a6465
3
+ size 6802
checkpoints/FINAL4_20260927/part_5/365m_dense16k_qkgainfp32_s42/checkpoint-22000-tamia/model.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6aecfe34697b05843a952e5263655daf90c831202f39a0f93e2fecb4700d7329
3
+ size 718046299
checkpoints/FINAL4_20260927/part_5/365m_dense16k_qkgainfp32_s42/checkpoint-22000-tamia/rng_state_rank0.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a1baa65d8b2ae1fa6f40523aa425ef947587299b9bae64d24b9fd6f56bbbe27b
3
+ size 14821
checkpoints/FINAL4_20260927/part_5/365m_dense16k_qkgainfp32_s42/checkpoint-22000-tamia/rng_state_rank1.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:39eaf0b28fae8a5f55ee8cd1d3ed6faad7411a9bc53c8517830ba190be3c514b
3
+ size 14757
checkpoints/FINAL4_20260927/part_5/365m_dense16k_qkgainfp32_s42/checkpoint-22000-tamia/rng_state_rank2.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:533bdfe5cf4e80da521f9e768df70535a9cfb919fa710c10fe8cb5cf196a8140
3
+ size 14821
checkpoints/FINAL4_20260927/part_5/365m_dense16k_qkgainfp32_s42/checkpoint-22000-tamia/rng_state_rank3.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9afacdc10ab864b577edf614e06695343a4ae8dbf802a36ade1d5f9eac9eb8fb
3
+ size 14821
checkpoints/FINAL4_20260927/part_5/365m_dense16k_qkgainfp32_s42/checkpoint-22000-tamia/train_config.json ADDED
@@ -0,0 +1,67 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model_path": "../qwen3_06b_base",
3
+ "model_dtype": "bfloat16",
4
+ "init_from_scratch": true,
5
+ "train_path": "../../data/fineweb_large/tokenized",
6
+ "val_path": null,
7
+ "max_seq_length": 16384,
8
+ "num_workers": 8,
9
+ "data_seed": 42,
10
+ "concatenate_seeded_8k": true,
11
+ "output_dir": "models/FINAL4_20260927/dense16k/365m_dense16k_qkgainfp32_s42/run_20261002_153011",
12
+ "resume_from_checkpoint": "models/FINAL4_20260927/dense16k/365m_dense16k_qkgainfp32_s42/run_20261002_153011/checkpoint-14000",
13
+ "force_new_run": false,
14
+ "num_epochs": 1,
15
+ "max_steps": 22000,
16
+ "per_device_batch_size": 1,
17
+ "gradient_accumulation_steps": 5,
18
+ "learning_rate": 0.0005,
19
+ "weight_decay": 0.01,
20
+ "warmup_steps": 750,
21
+ "lr_scheduler": "cosine",
22
+ "max_grad_norm": 1.0,
23
+ "save_steps": 1000,
24
+ "eval_steps": 250,
25
+ "eval_batches": 50,
26
+ "logging_steps": 100,
27
+ "seed": 42,
28
+ "verbose": false,
29
+ "checkpoint_activations": false,
30
+ "resume_micro_step_override": 70000,
31
+ "strategy": "ddp",
32
+ "mixed_precision": "bf16",
33
+ "wandb_enabled": false,
34
+ "wandb_project": "stochastic_sparsity",
35
+ "wandb_name": "365m_dense16k_qkgainfp32_s42",
36
+ "wandb_tags": [
37
+ "final4_20260927",
38
+ "dense16k",
39
+ "dense16k",
40
+ "seed42"
41
+ ],
42
+ "hidden_size": 768,
43
+ "intermediate_size": 3072,
44
+ "num_attention_heads": 6,
45
+ "num_key_value_heads": 2,
46
+ "num_hidden_layers": 28,
47
+ "head_dim": null,
48
+ "attention_dropout": null,
49
+ "rope_theta": 1000000.0,
50
+ "rope_train_position_scale_set": null,
51
+ "positional": "rope",
52
+ "qk_norm": "on",
53
+ "qk_gain_fp32": true,
54
+ "norm_gain_fp32": false,
55
+ "qk_gain_mode": "vector",
56
+ "attention_type": "dense",
57
+ "sparse_block_size": 128,
58
+ "sparse_num_random_blocks": 1,
59
+ "sparse_mask_pool_size": 1000,
60
+ "sparse_val_dense_compare": false,
61
+ "sparse_layer_indices": null,
62
+ "sparse_mask_cache_dir": null,
63
+ "sparse_mask_seed": 0,
64
+ "sparse_mixed_nrb_set": null,
65
+ "dense_mode": false,
66
+ "dense_step_intervals": null
67
+ }
checkpoints/FINAL4_20260927/part_5/365m_dense16k_qkgainfp32_s42/checkpoint-22000-tamia/trainer_state.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "global_step": 22000,
3
+ "epoch": 0,
4
+ "micro_step": 110000,
5
+ "sampler_seed": 42,
6
+ "seed": 42,
7
+ "world_size": 4,
8
+ "optimizer_state_file": "training_state.pt"
9
+ }
checkpoints/FINAL4_20260927/part_5/365m_dense16k_qkgainfp32_s42/checkpoint-22000-tamia/training_state.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9a534415a5f9fe505e9d46335953a4c78ea8780bc4406c9149e16344e2e6043a
3
+ size 1436134099