mate-local-models/image/gemlite/prism-ml/bonsai-image-binary-4B-gemlite-1bit/transformer-gemlite-int1/quantization_config.json

162 lines
No EOL
5.9 KiB
JSON

{
"format": "gemlite-int1-g128",
"bits": 1,
"group_size": 128,
"packing_bitwidth": 8,
"input_dtype": "fp16",
"output_dtype": "fp16",
"skip_patterns": [
"proj_out",
"x_embedder",
"context_embedder",
"time_text_embed",
"time_guidance_embed",
"norm_out",
"double_stream_modulation_img",
"double_stream_modulation_txt",
"single_stream_modulation"
],
"quantized_count": 100,
"skipped_count": 9,
"quantized_fqns": [
"transformer_blocks.0.attn.to_q",
"transformer_blocks.0.attn.to_k",
"transformer_blocks.0.attn.to_v",
"transformer_blocks.0.attn.add_q_proj",
"transformer_blocks.0.attn.add_k_proj",
"transformer_blocks.0.attn.add_v_proj",
"transformer_blocks.0.attn.to_add_out",
"transformer_blocks.0.attn.to_out.0",
"transformer_blocks.0.ff.linear_in",
"transformer_blocks.0.ff.linear_out",
"transformer_blocks.0.ff_context.linear_in",
"transformer_blocks.0.ff_context.linear_out",
"transformer_blocks.1.attn.to_q",
"transformer_blocks.1.attn.to_k",
"transformer_blocks.1.attn.to_v",
"transformer_blocks.1.attn.add_q_proj",
"transformer_blocks.1.attn.add_k_proj",
"transformer_blocks.1.attn.add_v_proj",
"transformer_blocks.1.attn.to_add_out",
"transformer_blocks.1.attn.to_out.0",
"transformer_blocks.1.ff.linear_in",
"transformer_blocks.1.ff.linear_out",
"transformer_blocks.1.ff_context.linear_in",
"transformer_blocks.1.ff_context.linear_out",
"transformer_blocks.2.attn.to_q",
"transformer_blocks.2.attn.to_k",
"transformer_blocks.2.attn.to_v",
"transformer_blocks.2.attn.add_q_proj",
"transformer_blocks.2.attn.add_k_proj",
"transformer_blocks.2.attn.add_v_proj",
"transformer_blocks.2.attn.to_add_out",
"transformer_blocks.2.attn.to_out.0",
"transformer_blocks.2.ff.linear_in",
"transformer_blocks.2.ff.linear_out",
"transformer_blocks.2.ff_context.linear_in",
"transformer_blocks.2.ff_context.linear_out",
"transformer_blocks.3.attn.to_q",
"transformer_blocks.3.attn.to_k",
"transformer_blocks.3.attn.to_v",
"transformer_blocks.3.attn.add_q_proj",
"transformer_blocks.3.attn.add_k_proj",
"transformer_blocks.3.attn.add_v_proj",
"transformer_blocks.3.attn.to_add_out",
"transformer_blocks.3.attn.to_out.0",
"transformer_blocks.3.ff.linear_in",
"transformer_blocks.3.ff.linear_out",
"transformer_blocks.3.ff_context.linear_in",
"transformer_blocks.3.ff_context.linear_out",
"transformer_blocks.4.attn.to_q",
"transformer_blocks.4.attn.to_k",
"transformer_blocks.4.attn.to_v",
"transformer_blocks.4.attn.add_q_proj",
"transformer_blocks.4.attn.add_k_proj",
"transformer_blocks.4.attn.add_v_proj",
"transformer_blocks.4.attn.to_add_out",
"transformer_blocks.4.attn.to_out.0",
"transformer_blocks.4.ff.linear_in",
"transformer_blocks.4.ff.linear_out",
"transformer_blocks.4.ff_context.linear_in",
"transformer_blocks.4.ff_context.linear_out",
"single_transformer_blocks.0.attn.to_qkv_mlp_proj",
"single_transformer_blocks.0.attn.to_out",
"single_transformer_blocks.1.attn.to_qkv_mlp_proj",
"single_transformer_blocks.1.attn.to_out",
"single_transformer_blocks.2.attn.to_qkv_mlp_proj",
"single_transformer_blocks.2.attn.to_out",
"single_transformer_blocks.3.attn.to_qkv_mlp_proj",
"single_transformer_blocks.3.attn.to_out",
"single_transformer_blocks.4.attn.to_qkv_mlp_proj",
"single_transformer_blocks.4.attn.to_out",
"single_transformer_blocks.5.attn.to_qkv_mlp_proj",
"single_transformer_blocks.5.attn.to_out",
"single_transformer_blocks.6.attn.to_qkv_mlp_proj",
"single_transformer_blocks.6.attn.to_out",
"single_transformer_blocks.7.attn.to_qkv_mlp_proj",
"single_transformer_blocks.7.attn.to_out",
"single_transformer_blocks.8.attn.to_qkv_mlp_proj",
"single_transformer_blocks.8.attn.to_out",
"single_transformer_blocks.9.attn.to_qkv_mlp_proj",
"single_transformer_blocks.9.attn.to_out",
"single_transformer_blocks.10.attn.to_qkv_mlp_proj",
"single_transformer_blocks.10.attn.to_out",
"single_transformer_blocks.11.attn.to_qkv_mlp_proj",
"single_transformer_blocks.11.attn.to_out",
"single_transformer_blocks.12.attn.to_qkv_mlp_proj",
"single_transformer_blocks.12.attn.to_out",
"single_transformer_blocks.13.attn.to_qkv_mlp_proj",
"single_transformer_blocks.13.attn.to_out",
"single_transformer_blocks.14.attn.to_qkv_mlp_proj",
"single_transformer_blocks.14.attn.to_out",
"single_transformer_blocks.15.attn.to_qkv_mlp_proj",
"single_transformer_blocks.15.attn.to_out",
"single_transformer_blocks.16.attn.to_qkv_mlp_proj",
"single_transformer_blocks.16.attn.to_out",
"single_transformer_blocks.17.attn.to_qkv_mlp_proj",
"single_transformer_blocks.17.attn.to_out",
"single_transformer_blocks.18.attn.to_qkv_mlp_proj",
"single_transformer_blocks.18.attn.to_out",
"single_transformer_blocks.19.attn.to_qkv_mlp_proj",
"single_transformer_blocks.19.attn.to_out"
],
"skipped": [
{
"fqn": "x_embedder",
"reason": "x_embedder"
},
{
"fqn": "context_embedder",
"reason": "context_embedder"
},
{
"fqn": "proj_out",
"reason": "proj_out"
},
{
"fqn": "time_guidance_embed.timestep_embedder.linear_1",
"reason": "time_guidance_embed"
},
{
"fqn": "time_guidance_embed.timestep_embedder.linear_2",
"reason": "time_guidance_embed"
},
{
"fqn": "double_stream_modulation_img.linear",
"reason": "double_stream_modulation_img"
},
{
"fqn": "double_stream_modulation_txt.linear",
"reason": "double_stream_modulation_txt"
},
{
"fqn": "single_stream_modulation.linear",
"reason": "single_stream_modulation"
},
{
"fqn": "norm_out.linear",
"reason": "norm_out"
}
],
"pack_seconds": 128.37
}