Views
No views yet
transformer-lab collection.| Field | Value |
|---|---|
| Repository | Pradheep1647/run_mqa-meetingbank-bs8-e20-fp32-19 |
| Attention | mqa |
| Dataset | meetingbank |
| Layers | 6 |
| Hidden size | 512 |
| Heads | 8 |
| Batch size | 8 |
| Epochs | 20 |
| Precision | fp32 |
| Checkpoint | meeting_model19.pt |
config.json, including model width, depth, sequence dimensions, and attention-specific settings.
loss_curve.csv.| File | Purpose |
|---|---|
meeting_model19.pt | PyTorch checkpoint containing model_state_dict, optimizer states, epoch, and global step. |
config.json | Training and architecture config converted from the Hydra run config. |
architecture.png | Architecture diagram generated from the saved model config, with block shapes and dimensions. |
tokenizer.json | MeetingBank transcript tokenizer alias for source inputs. |
transcript_tokenizer.json | Explicit MeetingBank transcript tokenizer. |
summary_tokenizer.json | MeetingBank summary tokenizer for target text. |
loss_curve.csv | TensorBoard train/loss scalar export. |
loss_curve.svg | Static training-loss plot generated from loss_curve.csv. |
transformers.AutoModel checkpoint. Use the repo-native builder to instantiate the architecture, then load the checkpoint state dict.1from pathlib import Path
2
3import torch
4from huggingface_hub import hf_hub_download
5from omegaconf import OmegaConf
6
7import src # registers components
8from src.model.builder import build_transformer
9
10repo_id = "Pradheep1647/run_mqa-meetingbank-bs8-e20-fp32-19"
11
12config_path = hf_hub_download(repo_id=repo_id, filename="config.json")
13checkpoint_path = hf_hub_download(repo_id=repo_id, filename="meeting_model19.pt")
14
15cfg = OmegaConf.load(config_path)
16model = build_transformer(cfg)
17
18state = torch.load(checkpoint_path, map_location="cpu")
19model.load_state_dict(state["model_state_dict"])
20model.eval()
21
22print(f"Loaded {repo_id} from {Path(checkpoint_path).name}")config.json as the source of truth for architecture parameters.