Instructions to use centml/wan2.2-t2v-a14b-diffusers-mlpinf-fp8 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use centml/wan2.2-t2v-a14b-diffusers-mlpinf-fp8 with Diffusers:
pip install -U diffusers transformers accelerate
import torch from diffusers import DiffusionPipeline # switch to "mps" for apple devices pipe = DiffusionPipeline.from_pretrained("centml/wan2.2-t2v-a14b-diffusers-mlpinf-fp8", dtype=torch.bfloat16, device_map="cuda") prompt = "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k" image = pipe(prompt).images[0] - Notebooks
- Google Colab
- Kaggle
add transformer_2/config.json
Browse files- transformer_2/config.json +66 -0
transformer_2/config.json
ADDED
|
@@ -0,0 +1,66 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_class_name": "WanTransformer3DModel",
|
| 3 |
+
"_diffusers_version": "0.37.1",
|
| 4 |
+
"_name_or_path": "/lustre/fsw/coreai_mlperf_inference/tinyinl/models/Wan2.2-T2V-A14B-Diffusers/transformer_2",
|
| 5 |
+
"added_kv_proj_dim": null,
|
| 6 |
+
"attention_head_dim": 128,
|
| 7 |
+
"cross_attn_norm": true,
|
| 8 |
+
"eps": 1e-06,
|
| 9 |
+
"ffn_dim": 13824,
|
| 10 |
+
"freq_dim": 256,
|
| 11 |
+
"image_dim": null,
|
| 12 |
+
"in_channels": 16,
|
| 13 |
+
"num_attention_heads": 40,
|
| 14 |
+
"num_layers": 40,
|
| 15 |
+
"out_channels": 16,
|
| 16 |
+
"patch_size": [
|
| 17 |
+
1,
|
| 18 |
+
2,
|
| 19 |
+
2
|
| 20 |
+
],
|
| 21 |
+
"pos_embed_seq_len": null,
|
| 22 |
+
"qk_norm": "rms_norm_across_heads",
|
| 23 |
+
"rope_max_seq_len": 1024,
|
| 24 |
+
"text_dim": 4096,
|
| 25 |
+
"quantization_config": {
|
| 26 |
+
"config_groups": {
|
| 27 |
+
"group_0": {
|
| 28 |
+
"input_activations": {
|
| 29 |
+
"dynamic": false,
|
| 30 |
+
"num_bits": 8,
|
| 31 |
+
"type": "float"
|
| 32 |
+
},
|
| 33 |
+
"weights": {
|
| 34 |
+
"dynamic": false,
|
| 35 |
+
"num_bits": 8,
|
| 36 |
+
"type": "float"
|
| 37 |
+
},
|
| 38 |
+
"targets": [
|
| 39 |
+
"Linear"
|
| 40 |
+
]
|
| 41 |
+
}
|
| 42 |
+
},
|
| 43 |
+
"ignore": [
|
| 44 |
+
"blocks.0*",
|
| 45 |
+
"blocks.1.*",
|
| 46 |
+
"blocks.2.*",
|
| 47 |
+
"blocks.37*",
|
| 48 |
+
"blocks.38*",
|
| 49 |
+
"blocks.39*",
|
| 50 |
+
"condition_embedder*",
|
| 51 |
+
"patch_embedding",
|
| 52 |
+
"proj_out"
|
| 53 |
+
],
|
| 54 |
+
"quant_algo": "FP8",
|
| 55 |
+
"kv_cache_scheme": {
|
| 56 |
+
"dynamic": false,
|
| 57 |
+
"num_bits": 8,
|
| 58 |
+
"type": "float"
|
| 59 |
+
},
|
| 60 |
+
"producer": {
|
| 61 |
+
"name": "modelopt",
|
| 62 |
+
"version": "0.46.0.dev170+gf479e7890"
|
| 63 |
+
},
|
| 64 |
+
"quant_method": "modelopt"
|
| 65 |
+
}
|
| 66 |
+
}
|