| { |
| "format": "split", |
| "model_version": "2.5.0", |
| "variant": "dev", |
| "tier": "q8", |
| "quantized": true, |
| "quantization_bits": 8, |
| "quantization_group_size": 64, |
| "components": [ |
| "transformer", |
| "text_encoder", |
| "connector", |
| "vae_decoder", |
| "vae_encoder", |
| "diffusion_vae_encoder", |
| "vae_diffusion_decoder", |
| "audio_vae", |
| "vocoder", |
| "spatial_upsampler", |
| "temporal_upsampler", |
| "duration_head" |
| ], |
| "component_detail": [ |
| { |
| "name": "transformer", |
| "file": "transformer.safetensors", |
| "tensors": 6779, |
| "quantized_linears": 1344, |
| "dense_float_tensors": 2747, |
| "passthrough_tensors": 0, |
| "bytes": 20612172436 |
| }, |
| { |
| "name": "text_encoder", |
| "file": "text_encoder.safetensors", |
| "tensors": 1338, |
| "quantized_linears": 328, |
| "dense_float_tensors": 349, |
| "passthrough_tensors": 5, |
| "bytes": 13733108668 |
| }, |
| { |
| "name": "connector", |
| "file": "connector.safetensors", |
| "tensors": 458, |
| "quantized_linears": 98, |
| "dense_float_tensors": 164, |
| "passthrough_tensors": 0, |
| "bytes": 3373318161 |
| }, |
| { |
| "name": "vae_decoder", |
| "file": "vae_decoder.safetensors", |
| "tensors": 86, |
| "quantized_linears": 0, |
| "dense_float_tensors": 86, |
| "passthrough_tensors": 0, |
| "bytes": 814384925, |
| "dense_reason": "no-linear-weights", |
| "dense_reason_detail": "no rank-2 Linear weights: every weight is a convolution kernel, a norm or a per-channel statistic, and MLX has no quantized convolution" |
| }, |
| { |
| "name": "vae_encoder", |
| "file": "vae_encoder.safetensors", |
| "tensors": 86, |
| "quantized_linears": 0, |
| "dense_float_tensors": 86, |
| "passthrough_tensors": 0, |
| "bytes": 637919624, |
| "dense_reason": "no-linear-weights", |
| "dense_reason_detail": "no rank-2 Linear weights: every weight is a convolution kernel, a norm or a per-channel statistic, and MLX has no quantized convolution" |
| }, |
| { |
| "name": "diffusion_vae_encoder", |
| "file": "diffusion_vae_encoder.safetensors", |
| "tensors": 86, |
| "quantized_linears": 0, |
| "dense_float_tensors": 86, |
| "passthrough_tensors": 0, |
| "bytes": 637919563, |
| "dense_reason": "no-linear-weights", |
| "dense_reason_detail": "no rank-2 Linear weights: every weight is a convolution kernel, a norm or a per-channel statistic, and MLX has no quantized convolution" |
| }, |
| { |
| "name": "vae_diffusion_decoder", |
| "file": "vae_diffusion_decoder.safetensors", |
| "tensors": 586, |
| "quantized_linears": 137, |
| "dense_float_tensors": 175, |
| "passthrough_tensors": 0, |
| "bytes": 443456318 |
| }, |
| { |
| "name": "audio_vae", |
| "file": "audio_vae.safetensors", |
| "tensors": 58, |
| "quantized_linears": 0, |
| "dense_float_tensors": 58, |
| "passthrough_tensors": 0, |
| "bytes": 63873569, |
| "dense_reason": "no-linear-weights", |
| "dense_reason_detail": "no rank-2 Linear weights: every weight is a convolution kernel, a norm or a per-channel statistic, and MLX has no quantized convolution" |
| }, |
| { |
| "name": "vocoder", |
| "file": "vocoder.safetensors", |
| "tensors": 1227, |
| "quantized_linears": 0, |
| "dense_float_tensors": 1227, |
| "passthrough_tensors": 0, |
| "bytes": 258339943, |
| "dense_reason": "no-linear-weights", |
| "dense_reason_detail": "no rank-2 Linear weights: every weight is a convolution kernel, a norm or a per-channel statistic, and MLX has no quantized convolution" |
| }, |
| { |
| "name": "spatial_upsampler", |
| "file": "spatial_upsampler.safetensors", |
| "tensors": 72, |
| "quantized_linears": 0, |
| "dense_float_tensors": 72, |
| "passthrough_tensors": 0, |
| "bytes": 995778754, |
| "dense_reason": "no-linear-weights", |
| "dense_reason_detail": "no rank-2 Linear weights: every weight is a convolution kernel, a norm or a per-channel statistic, and MLX has no quantized convolution" |
| }, |
| { |
| "name": "temporal_upsampler", |
| "file": "temporal_upsampler.safetensors", |
| "tensors": 72, |
| "quantized_linears": 0, |
| "dense_float_tensors": 72, |
| "passthrough_tensors": 0, |
| "bytes": 261944030, |
| "dense_reason": "no-linear-weights", |
| "dense_reason_detail": "no rank-2 Linear weights: every weight is a convolution kernel, a norm or a per-channel statistic, and MLX has no quantized convolution" |
| }, |
| { |
| "name": "duration_head", |
| "file": "duration_head.safetensors", |
| "tensors": 15, |
| "quantized_linears": 0, |
| "dense_float_tensors": 15, |
| "passthrough_tensors": 0, |
| "bytes": 3843697, |
| "dense_reason": "no-mlx-port", |
| "dense_reason_detail": "this crate has no MLX port that can run these weights yet (sc-18777); quantizing them would ship weights no test can exercise" |
| } |
| ], |
| "total_bytes": 41836059688 |
| } |
