kentang1998
commited on
Commit
·
748f83b
1
Parent(s):
b9e0976
[Major] Add ckpt
Browse files- llm/config.json +53 -0
- llm/ema_model.bin +3 -0
- llm/pytorch_model.bin +3 -0
- llm/trainer_state.json +21 -0
- tokenizer/config.json +54 -0
- tokenizer/pytorch_model.bin +3 -0
- tokenizer/trainer_state.json +20 -0
llm/config.json
ADDED
@@ -0,0 +1,53 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"architectures": [
|
3 |
+
"HARTForT2I"
|
4 |
+
],
|
5 |
+
"attn_drop_rate": 0.0,
|
6 |
+
"attn_l2_norm": false,
|
7 |
+
"attn_type": "llama",
|
8 |
+
"cond_drop_rate": 0.1,
|
9 |
+
"context_dim": 1536,
|
10 |
+
"context_norm_scale": 1.0,
|
11 |
+
"context_token": 300,
|
12 |
+
"depth": 24,
|
13 |
+
"diff_depth": 6,
|
14 |
+
"diff_width": 1024,
|
15 |
+
"diffusion_batch_mul": 4,
|
16 |
+
"diffusion_head_repeats": 1,
|
17 |
+
"disable_aln": true,
|
18 |
+
"drop_path_rate": 0.10000000000000002,
|
19 |
+
"drop_rate": 0.0,
|
20 |
+
"embed_dim": 1536,
|
21 |
+
"flash_if_available": true,
|
22 |
+
"fused_if_available": true,
|
23 |
+
"mlp_ratio": 4.0,
|
24 |
+
"mlp_type": "llama",
|
25 |
+
"model_type": "hart_transformer_t2i",
|
26 |
+
"norm_eps": 1e-06,
|
27 |
+
"num_heads": 24,
|
28 |
+
"num_sampling_steps": "8",
|
29 |
+
"patch_nums": [
|
30 |
+
1,
|
31 |
+
2,
|
32 |
+
3,
|
33 |
+
4,
|
34 |
+
5,
|
35 |
+
7,
|
36 |
+
9,
|
37 |
+
12,
|
38 |
+
16,
|
39 |
+
21,
|
40 |
+
27,
|
41 |
+
36,
|
42 |
+
48,
|
43 |
+
64
|
44 |
+
],
|
45 |
+
"sampler": "iddpm",
|
46 |
+
"sep_aln_pooling_mode": "max",
|
47 |
+
"shared_aln": false,
|
48 |
+
"torch_dtype": "float32",
|
49 |
+
"transformers_version": "4.42.2",
|
50 |
+
"use_context_norm": true,
|
51 |
+
"use_cross_attn": false,
|
52 |
+
"use_timestep_embed": true
|
53 |
+
}
|
llm/ema_model.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:83ee6130434089cd570929eeadae7a24708cf99f29b08ddb67e4175126cbacb1
|
3 |
+
size 3310193530
|
llm/pytorch_model.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:b0b07632c4efd2ce1b34c002287410ec7a6ed321ba23e79d6a6c2d080ed4f64b
|
3 |
+
size 3310175358
|
llm/trainer_state.json
ADDED
@@ -0,0 +1,21 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"best_metric": 5.495192527770996,
|
3 |
+
"best_model_checkpoint": "ckpts/transformer_hybrid_t2i_haotian/r1024from512_MJ36M_730M+maskgit+diff+qwenvl+norm+prefix-prompting_ctxmask_8nodes_small_batch_grad_ckpting154k+66k/checkpoint-30000",
|
4 |
+
"epoch": 0.160668380462606,
|
5 |
+
"eval_steps": 500,
|
6 |
+
"global_step": 30000,
|
7 |
+
"is_hyper_param_search": false,
|
8 |
+
"is_local_process_zero": true,
|
9 |
+
"is_world_process_zero": true,
|
10 |
+
"log_history": [],
|
11 |
+
"logging_steps": 10,
|
12 |
+
"max_steps": 0,
|
13 |
+
"num_input_tokens_seen": 0,
|
14 |
+
"num_train_epochs": 50,
|
15 |
+
"save_steps": 200,
|
16 |
+
"stateful_callbacks": {},
|
17 |
+
"total_flos": 0,
|
18 |
+
"train_batch_size": null,
|
19 |
+
"trial_name": null,
|
20 |
+
"trial_params": null
|
21 |
+
}
|
tokenizer/config.json
ADDED
@@ -0,0 +1,54 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"architectures": [
|
3 |
+
"HARTAutoEncoderWithDisc"
|
4 |
+
],
|
5 |
+
"beta": 0.25,
|
6 |
+
"ch": 160,
|
7 |
+
"ch_mult": [
|
8 |
+
1,
|
9 |
+
1,
|
10 |
+
2,
|
11 |
+
2,
|
12 |
+
4
|
13 |
+
],
|
14 |
+
"default_qresi_counts": 0,
|
15 |
+
"disable_quant_resi": false,
|
16 |
+
"double_decoder": false,
|
17 |
+
"dropout": 0.0,
|
18 |
+
"freeze_codebook_for_hybrid": true,
|
19 |
+
"hybrid": false,
|
20 |
+
"levels": [
|
21 |
+
8,
|
22 |
+
8,
|
23 |
+
8,
|
24 |
+
6,
|
25 |
+
5
|
26 |
+
],
|
27 |
+
"model_type": "hart_autoencoder_with_disc",
|
28 |
+
"quant_conv_ks": 3,
|
29 |
+
"quant_resi": 0.5,
|
30 |
+
"quantizer_type": "var_hybrid",
|
31 |
+
"share_quant_resi": 4,
|
32 |
+
"test_mode": false,
|
33 |
+
"torch_dtype": "float32",
|
34 |
+
"transformers_version": "4.36.2",
|
35 |
+
"using_znorm": false,
|
36 |
+
"v_patch_nums": [
|
37 |
+
1,
|
38 |
+
2,
|
39 |
+
3,
|
40 |
+
4,
|
41 |
+
5,
|
42 |
+
7,
|
43 |
+
9,
|
44 |
+
12,
|
45 |
+
16,
|
46 |
+
21,
|
47 |
+
27,
|
48 |
+
36,
|
49 |
+
48,
|
50 |
+
64
|
51 |
+
],
|
52 |
+
"vocab_size": 4096,
|
53 |
+
"z_channels": 32
|
54 |
+
}
|
tokenizer/pytorch_model.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:550cbc8abf5a9a6ce21404a5432ffb9ef7cdbde3b43468b081e361ff7d5206ef
|
3 |
+
size 447207858
|
tokenizer/trainer_state.json
ADDED
@@ -0,0 +1,20 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"best_metric": 0.33839765191078186,
|
3 |
+
"best_model_checkpoint": "ckpts/vae_hybrid/open-images-disc0.2-f16c32-2nodes-hybrid-resume-discrete-single-decoder-r1024/checkpoint-5000",
|
4 |
+
"epoch": 0.0195355233957447,
|
5 |
+
"eval_steps": 1000,
|
6 |
+
"global_step": 10000,
|
7 |
+
"is_hyper_param_search": false,
|
8 |
+
"is_local_process_zero": true,
|
9 |
+
"is_world_process_zero": true,
|
10 |
+
"log_history": [],
|
11 |
+
"logging_steps": 10,
|
12 |
+
"max_steps": 0,
|
13 |
+
"num_input_tokens_seen": 0,
|
14 |
+
"num_train_epochs": 1,
|
15 |
+
"save_steps": 500,
|
16 |
+
"total_flos": 0,
|
17 |
+
"train_batch_size": null,
|
18 |
+
"trial_name": null,
|
19 |
+
"trial_params": null
|
20 |
+
}
|