Attribute as quantised child of Adobe Joint SpeakerID; drop GitHub fork links

README.md

@@ -3,6 +3,7 @@ license: other
 license_name: adobe-research-license
 license_link: https://github.com/adobe-research/speaker-identification/blob/main/LICENSE.md
 library_name: pytorch
+base_model: FacebookAI/roberta-large
 tags:
   - speaker-identification
   - quantization
@@ -17,20 +18,33 @@ INT8 weight-only quantisation of the Joint Speaker Identifier from
 [adobe-research/speaker-identification](https://github.com/adobe-research/speaker-identification)
 ([Interspeech 2024](https://arxiv.org/abs/2407.12094)).
 
+This checkpoint is a **quantised child** of Adobe Research’s original Joint
+Speaker Identifier (FP32). The architecture and trained weights are Adobe’s;
+only the Linear weights were packed to INT8 (group size 64). No additional
+training.
+
 **Joint INT8 — 78.87 / 0.00 / 63.28 / 67.60 / 18.34 / 472 MB**
 
 | Metric | Value |
 |---|---|
 | Precision | **78.87** |
-| Δ vs FP32 | **0.00** |
+| Δ vs FP32 parent | **0.00** |
 | F1 | 63.28 |
 | Accuracy | 67.60 |
 | Throughput | 18.34 examples/s (Apple M3 Pro, MPS, batch 2) |
-| In-memory size | 472 MB (FP32: 1633 MB) |
+| In-memory size | 472 MB (FP32 parent: 1633 MB) |
+
+Same precision, F1, and accuracy as the FP32 Joint parent at about 3.5× smaller weights.
 
-Same precision, F1, and accuracy as the FP32 Joint baseline at about 3.5× smaller weights.
+## Parent model
 
-Code and eval tables: [HugoMarchant/speaker-identification-quantisation](https://github.com/HugoMarchant/speaker-identification-quantisation).
+| | |
+|---|---|
+| Parent | Joint Speaker Identifier (FP32), Adobe Research |
+| Original weights | `logs/mediasum-joint/best-model.mdl` in [adobe-research/speaker-identification](https://github.com/adobe-research/speaker-identification) |
+| Paper | [Identifying Speakers in Dialogue Transcripts: A Text-based Approach Using Pretrained Language Models](https://arxiv.org/abs/2407.12094) (Interspeech 2024) |
+| Backbone | [FacebookAI/roberta-large](https://huggingface.co/FacebookAI/roberta-large) |
+| Relation | weight-only INT8 quantisation of the Adobe Joint checkpoint |
 
 ## Files
 
@@ -41,16 +55,10 @@ Code and eval tables: [HugoMarchant/speaker-identification-quantisation](https:/
 
 ```python
 from huggingface_hub import hf_hub_download
-from device_utils import apply_cuda_shims
-apply_cuda_shims()
-from model_io import load_quantized_bundle
 
 path = hf_hub_download("hmarchant/speaker-id-joint-int8", "model.pt")
-model, config, tokenizer, qcfg, info, device = load_quantized_bundle(path)
 ```
 
-Requires the companion GitHub repo (custom `JointSpeakerIdentifier` + `WeightOnlyQuantizedLinear`).
-
 ## License
 
 Derived from Adobe Research Speaker Identification. The