Hugging Face's logo

topabaem
/
LTX-2.5-Text-Encoder-4bit-8GB

Image-to-Video
quantization
4-bit precision
int4
nvfp4
gptq
awq
text-encoder
text-to-video
video-to-video
ltx-video
low-vram
comfyui
Model card Files
xet
 Community
1
LTX-2.5-Text-Encoder-4bit-8GB
File size: 10,482 Bytes
d7d6ccb
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71

Loading weights:   0%|          | 0/677 [00:00<?, ?it/s]
Loading weights:   0%|          | 2/677 [00:00<02:26,  4.62it/s]
Loading weights:   2%|▏         | 13/677 [00:00<00:21, 30.56it/s]
Loading weights:   3%|▎         | 20/677 [00:00<00:17, 38.10it/s]
Loading weights:   5%|▌         | 35/677 [00:00<00:09, 64.30it/s]
Loading weights:   7%|▋         | 49/677 [00:00<00:07, 80.65it/s]
Loading weights:   9%|▉         | 62/677 [00:01<00:06, 90.20it/s]
Loading weights:  11%|█▏        | 77/677 [00:01<00:05, 100.78it/s]
Loading weights:  13%|█▎        | 89/677 [00:01<00:05, 98.60it/s] 
Loading weights:  15%|█▌        | 104/677 [00:01<00:05, 107.23it/s]
Loading weights:  17%|█▋        | 117/677 [00:01<00:05, 108.29it/s]
Loading weights:  19%|█▉        | 132/677 [00:01<00:04, 113.28it/s]
Loading weights:  22%|██▏       | 146/677 [00:01<00:04, 115.12it/s]
Loading weights:  24%|██▎       | 160/677 [00:01<00:04, 117.13it/s]
Loading weights:  26%|██▌       | 173/677 [00:01<00:04, 113.22it/s]
Loading weights:  27%|██▋       | 186/677 [00:02<00:04, 113.22it/s]
Loading weights:  30%|██▉       | 200/677 [00:02<00:04, 115.41it/s]
Loading weights:  32%|███▏      | 215/677 [00:02<00:03, 119.73it/s]
Loading weights:  34%|███▍      | 229/677 [00:02<00:03, 119.61it/s]
Loading weights:  36%|███▌      | 243/677 [00:02<00:03, 119.16it/s]
Loading weights:  38%|███▊      | 256/677 [00:02<00:03, 114.45it/s]
Loading weights:  40%|███▉      | 269/677 [00:02<00:03, 113.27it/s]
Loading weights:  42%|████▏     | 284/677 [00:02<00:03, 117.67it/s]
Loading weights:  44%|████▍     | 298/677 [00:03<00:03, 118.56it/s]
Loading weights:  46%|████▌     | 312/677 [00:03<00:03, 119.22it/s]
Loading weights:  48%|████▊     | 326/677 [00:03<00:02, 118.33it/s]
Loading weights:  50%|█████     | 339/677 [00:03<00:02, 114.21it/s]
Loading weights:  52%|█████▏    | 352/677 [00:03<00:02, 113.79it/s]
Loading weights:  54%|█████▍    | 366/677 [00:03<00:02, 115.77it/s]
Loading weights:  56%|█████▌    | 380/677 [00:03<00:02, 117.48it/s]
Loading weights:  58%|█████▊    | 395/677 [00:03<00:02, 121.10it/s]
Loading weights:  60%|██████    | 408/677 [00:03<00:02, 118.14it/s]
Loading weights:  62%|██████▏   | 421/677 [00:04<00:02, 113.96it/s]
Loading weights:  64%|██████▍   | 435/677 [00:04<00:02, 116.14it/s]
Loading weights:  66%|██████▋   | 450/677 [00:04<00:01, 119.95it/s]
Loading weights:  68%|██████▊   | 463/677 [00:04<00:01, 117.41it/s]
Loading weights:  71%|███████   | 478/677 [00:04<00:01, 121.13it/s]
Loading weights:  73%|███████▎  | 491/677 [00:04<00:01, 118.51it/s]
Loading weights:  74%|███████▍  | 503/677 [00:04<00:01, 111.44it/s]
Loading weights:  77%|███████▋  | 518/677 [00:04<00:01, 117.00it/s]
Loading weights:  79%|███████▊  | 532/677 [00:05<00:01, 118.05it/s]
Loading weights:  81%|████████  | 547/677 [00:05<00:01, 121.54it/s]
Loading weights:  83%|████████▎ | 561/677 [00:05<00:00, 121.76it/s]
Loading weights:  85%|████████▍ | 574/677 [00:05<00:00, 118.81it/s]
Loading weights:  87%|████████▋ | 588/677 [00:05<00:00, 116.75it/s]
Loading weights:  89%|████████▉ | 601/677 [00:05<00:00, 115.47it/s]
Loading weights:  91%|█████████ | 615/677 [00:05<00:00, 117.17it/s]
Loading weights:  93%|█████████▎| 630/677 [00:05<00:00, 120.81it/s]
Loading weights:  95%|█████████▌| 644/677 [00:05<00:00, 120.14it/s]
Loading weights:  97%|█████████▋| 658/677 [00:06<00:00, 119.24it/s]
Loading weights: 100%|██████████| 677/677 [00:06<00:00, 110.23it/s]
[transformers] Gemma4UnifiedModel LOAD REPORT from: /content/ltxwork/encoder-hf
Key                                                    | Status     |  | 
-------------------------------------------------------+------------+--+-
text_embedding_projection.video_aggregate_embed.weight | UNEXPECTED |  | 
hf_asset__processor_config.json                        | UNEXPECTED |  | 
hf_asset__chat_template.jinja                          | UNEXPECTED |  | 
hf_asset__generation_config.json                       | UNEXPECTED |  | 
text_embedding_projection.audio_aggregate_embed.weight | UNEXPECTED |  | 
hf_asset__tokenizer_config.json                        | UNEXPECTED |  | 
tokenizer_json                                         | UNEXPECTED |  | 
text_embedding_projection.audio_aggregate_embed.bias   | UNEXPECTED |  | 
text_embedding_projection.video_aggregate_embed.bias   | UNEXPECTED |  | 

Notes:
- UNEXPECTED:	can be ignored when loading from different task/architecture; not ok if you expect identical arch.
captured contexts for 33 distinct lengths
calibration: 146 batches, 166028 tokens
wrote 22 non-decoder tensors
layer  0 [slid] 7 quantized, worst diagonal output relL2 0.1508, worst ||Q||/||W|| 4.216, dead columns 0
layer  1 [slid] 7 quantized, worst diagonal output relL2 0.1282, worst ||Q||/||W|| 1.058, dead columns 0
layer  2 [slid] 7 quantized, worst diagonal output relL2 0.1153, worst ||Q||/||W|| 1.006, dead columns 0
layer  3 [slid] 7 quantized, worst diagonal output relL2 0.1183, worst ||Q||/||W|| 1.010, dead columns 0
layer  4 [slid] 7 quantized, worst diagonal output relL2 0.1150, worst ||Q||/||W|| 1.007, dead columns 0
layer  5 [full] 5 quantized, worst diagonal output relL2 0.1151, worst ||Q||/||W|| 1.008, dead columns 0
layer  6 [slid] 7 quantized, worst diagonal output relL2 0.1184, worst ||Q||/||W|| 1.009, dead columns 0
layer  7 [slid] 7 quantized, worst diagonal output relL2 0.1213, worst ||Q||/||W|| 1.008, dead columns 0
layer  8 [slid] 7 quantized, worst diagonal output relL2 0.1126, worst ||Q||/||W|| 1.014, dead columns 0
layer  9 [slid] 7 quantized, worst diagonal output relL2 0.1102, worst ||Q||/||W|| 1.015, dead columns 0
layer 10 [slid] 7 quantized, worst diagonal output relL2 0.1134, worst ||Q||/||W|| 1.120, dead columns 0
layer 11 [full] 5 quantized, worst diagonal output relL2 0.1114, worst ||Q||/||W|| 1.010, dead columns 0
layer 12 [slid] 7 quantized, worst diagonal output relL2 0.1198, worst ||Q||/||W|| 1.079, dead columns 0
layer 13 [slid] 7 quantized, worst diagonal output relL2 0.1177, worst ||Q||/||W|| 1.069, dead columns 0
layer 14 [slid] 7 quantized, worst diagonal output relL2 0.1178, worst ||Q||/||W|| 1.038, dead columns 0
layer 15 [slid] 7 quantized, worst diagonal output relL2 0.1137, worst ||Q||/||W|| 1.027, dead columns 0
layer 16 [slid] 7 quantized, worst diagonal output relL2 0.1084, worst ||Q||/||W|| 1.039, dead columns 0
layer 17 [full] 5 quantized, worst diagonal output relL2 0.1107, worst ||Q||/||W|| 1.018, dead columns 0
layer 18 [slid] 7 quantized, worst diagonal output relL2 0.1135, worst ||Q||/||W|| 1.008, dead columns 0
layer 19 [slid] 7 quantized, worst diagonal output relL2 0.1113, worst ||Q||/||W|| 1.006, dead columns 0
layer 20 [slid] 7 quantized, worst diagonal output relL2 0.1156, worst ||Q||/||W|| 1.012, dead columns 0
layer 21 [slid] 7 quantized, worst diagonal output relL2 0.1163, worst ||Q||/||W|| 1.012, dead columns 0
layer 22 [slid] 7 quantized, worst diagonal output relL2 0.1172, worst ||Q||/||W|| 1.008, dead columns 0
layer 23 [full] 5 quantized, worst diagonal output relL2 0.1159, worst ||Q||/||W|| 1.011, dead columns 0
layer 24 [slid] 7 quantized, worst diagonal output relL2 0.1161, worst ||Q||/||W|| 1.007, dead columns 0
layer 25 [slid] 7 quantized, worst diagonal output relL2 0.1129, worst ||Q||/||W|| 1.006, dead columns 0
layer 26 [slid] 7 quantized, worst diagonal output relL2 0.1176, worst ||Q||/||W|| 1.006, dead columns 0
layer 27 [slid] 7 quantized, worst diagonal output relL2 0.1144, worst ||Q||/||W|| 1.006, dead columns 0
layer 28 [slid] 7 quantized, worst diagonal output relL2 0.1217, worst ||Q||/||W|| 1.007, dead columns 0
layer 29 [full] 5 quantized, worst diagonal output relL2 0.1132, worst ||Q||/||W|| 1.007, dead columns 0
layer 30 [slid] 7 quantized, worst diagonal output relL2 0.1114, worst ||Q||/||W|| 1.005, dead columns 0
layer 31 [slid] 7 quantized, worst diagonal output relL2 0.1075, worst ||Q||/||W|| 1.005, dead columns 0
layer 32 [slid] 7 quantized, worst diagonal output relL2 0.1133, worst ||Q||/||W|| 1.006, dead columns 0
layer 33 [slid] 7 quantized, worst diagonal output relL2 0.1122, worst ||Q||/||W|| 1.012, dead columns 0
layer 34 [slid] 7 quantized, worst diagonal output relL2 0.1235, worst ||Q||/||W|| 1.004, dead columns 0
layer 35 [full] 5 quantized, worst diagonal output relL2 0.1177, worst ||Q||/||W|| 1.008, dead columns 0
layer 36 [slid] 7 quantized, worst diagonal output relL2 0.1174, worst ||Q||/||W|| 1.006, dead columns 0
layer 37 [slid] 7 quantized, worst diagonal output relL2 0.1162, worst ||Q||/||W|| 1.006, dead columns 0
layer 38 [slid] 7 quantized, worst diagonal output relL2 0.1146, worst ||Q||/||W|| 1.006, dead columns 0
layer 39 [slid] 7 quantized, worst diagonal output relL2 0.1153, worst ||Q||/||W|| 1.004, dead columns 0
layer 40 [slid] 7 quantized, worst diagonal output relL2 0.1146, worst ||Q||/||W|| 1.004, dead columns 0
layer 41 [full] 5 quantized, worst diagonal output relL2 0.1136, worst ||Q||/||W|| 1.004, dead columns 0
layer 42 [slid] 7 quantized, worst diagonal output relL2 0.1147, worst ||Q||/||W|| 1.004, dead columns 0
layer 43 [slid] 7 quantized, worst diagonal output relL2 0.1136, worst ||Q||/||W|| 1.003, dead columns 0
layer 44 [slid] 7 quantized, worst diagonal output relL2 0.1101, worst ||Q||/||W|| 1.004, dead columns 0
layer 45 [slid] 7 quantized, worst diagonal output relL2 0.1063, worst ||Q||/||W|| 1.005, dead columns 0
layer 46 [slid] 7 quantized, worst diagonal output relL2 0.1435, worst ||Q||/||W|| 1.008, dead columns 0
layer 47 [full] 5 quantized, worst diagonal output relL2 0.0870, worst ||Q||/||W|| 1.007, dead columns 0
dead-guard bypass eps 1e-10 awq-floor 1e-05 restore-gain-max 64 damping median-active awq-group bounded
  worst ||Q||/||W|| 4.216 | columns zeroed 0 | max restore gain 19.00 | channels exact-zero 5 near-dead 967 | worst damping escalation 10x
wrote /content/builds/A3.packed.safetensors (8.464 GB packed, 19945148 zero-sign collapses)