Update README.md
Browse files
README.md
CHANGED
|
@@ -20,7 +20,6 @@ Pre-trained Megatron-format checkpoints for [MUG-V 10B](https://github.com/Shope
|
|
| 20 |
|
| 21 |
**Torch Distributed Checkpoint** - Flexible parallelism support
|
| 22 |
|
| 23 |
-
- **Size**: ~64 GB
|
| 24 |
- **Format**: Torch Distributed (`.distcp`)
|
| 25 |
- **Parallelism**: Can be loaded with **any TP/PP configuration**
|
| 26 |
- **Use Case**: Production training, flexible distributed setup
|
|
@@ -33,7 +32,6 @@ huggingface-cli download MUG-V/MUG-V-training --local-dir ./checkpoints --includ
|
|
| 33 |
|
| 34 |
**Torch Format (Legacy)** - Fixed TP=4
|
| 35 |
|
| 36 |
-
- **Size**: ~64 GB
|
| 37 |
- **Format**: Torch format (`mp_rank_XX/model_optim_rng.pt`)
|
| 38 |
- **Parallelism**: Must be loaded with **TP=4**
|
| 39 |
- **Use Case**: Fixed TP setup or conversion to Torch Distributed
|
|
|
|
| 20 |
|
| 21 |
**Torch Distributed Checkpoint** - Flexible parallelism support
|
| 22 |
|
|
|
|
| 23 |
- **Format**: Torch Distributed (`.distcp`)
|
| 24 |
- **Parallelism**: Can be loaded with **any TP/PP configuration**
|
| 25 |
- **Use Case**: Production training, flexible distributed setup
|
|
|
|
| 32 |
|
| 33 |
**Torch Format (Legacy)** - Fixed TP=4
|
| 34 |
|
|
|
|
| 35 |
- **Format**: Torch format (`mp_rank_XX/model_optim_rng.pt`)
|
| 36 |
- **Parallelism**: Must be loaded with **TP=4**
|
| 37 |
- **Use Case**: Fixed TP setup or conversion to Torch Distributed
|