2023-06-16 09:52:51 +00:00
|
|
|
import pytest
|
|
|
|
|
import torch
|
|
|
|
|
|
|
|
|
|
from petals.server.block_utils import resolve_block_dtype
|
Add LLaMA support (#323)
This PR:
1. **Abolishes the model conversion procedure.** Now, models are downloaded directly from original repositories like https://huggingface.co/bigscience/bloom. Servers download only shards with blocks to be hosted, and clients download only shards with input/output embeddings and layernorms.
- BLOOM is loaded from `bigscience/bloom`, but we use the DHT prefix `bigscience/bloom-petals` for backward compatibility. Same with smaller BLOOMs and BLOOMZ.
- LLaMA can be loaded from any repo like `username/llama-65b-hf`, but we use the DHT prefix `llama-65b-hf` (without the username) to accomodate blocks from different repos (there're a few of them with minor differences, such as `Llama` vs. `LLaMA` in the class name).
2. **Refactors the client to generalize it for multiple models.** Now, we have `petals.models` packages that contain model-specific code (e.g. `petals.models.bloom`, `petals.models.llama`). General code (e.g. CPU-efficient LM head, p-tuning) is kept in `petals.client`.
3. **Introduces** `WrappedLlamaBlock`, `DistributedLlamaConfig`, `DistributedLlamaForCausalLM`, `DistributedLlamaForSequenceClassification`, and `DistributedLlamaModel` compatible with Petals functionality (p-tuning, adapters, etc.).
4. **Introduces** `AutoDistributedConfig` that automatically chooses the correct config class (`DistributedLlamaConfig` or `DistributedBloomConfig`). The refactored configs contain all model-specific info for both clients and servers.
Upgrade instructions:
- Remove disk caches for blocks in old (converted) format to save disk space. That is, remove `~/.cache/petals/model--bigscience--bloom-petals` and `~/.cache/petals/model--bigscience--bloomz-petals` directories (if present).
2023-06-23 11:46:10 +00:00
|
|
|
from petals.server.from_pretrained import load_pretrained_block
|
|
|
|
|
from petals.utils.auto_config import AutoDistributedConfig
|
2023-06-16 09:52:51 +00:00
|
|
|
from test_utils import MODEL_NAME
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
@pytest.mark.forked
|
|
|
|
|
@pytest.mark.parametrize("torch_dtype", [torch.float32, torch.float16, "auto"])
|
Add LLaMA support (#323)
This PR:
1. **Abolishes the model conversion procedure.** Now, models are downloaded directly from original repositories like https://huggingface.co/bigscience/bloom. Servers download only shards with blocks to be hosted, and clients download only shards with input/output embeddings and layernorms.
- BLOOM is loaded from `bigscience/bloom`, but we use the DHT prefix `bigscience/bloom-petals` for backward compatibility. Same with smaller BLOOMs and BLOOMZ.
- LLaMA can be loaded from any repo like `username/llama-65b-hf`, but we use the DHT prefix `llama-65b-hf` (without the username) to accomodate blocks from different repos (there're a few of them with minor differences, such as `Llama` vs. `LLaMA` in the class name).
2. **Refactors the client to generalize it for multiple models.** Now, we have `petals.models` packages that contain model-specific code (e.g. `petals.models.bloom`, `petals.models.llama`). General code (e.g. CPU-efficient LM head, p-tuning) is kept in `petals.client`.
3. **Introduces** `WrappedLlamaBlock`, `DistributedLlamaConfig`, `DistributedLlamaForCausalLM`, `DistributedLlamaForSequenceClassification`, and `DistributedLlamaModel` compatible with Petals functionality (p-tuning, adapters, etc.).
4. **Introduces** `AutoDistributedConfig` that automatically chooses the correct config class (`DistributedLlamaConfig` or `DistributedBloomConfig`). The refactored configs contain all model-specific info for both clients and servers.
Upgrade instructions:
- Remove disk caches for blocks in old (converted) format to save disk space. That is, remove `~/.cache/petals/model--bigscience--bloom-petals` and `~/.cache/petals/model--bigscience--bloomz-petals` directories (if present).
2023-06-23 11:46:10 +00:00
|
|
|
def test_block_dtype(torch_dtype):
|
|
|
|
|
config = AutoDistributedConfig.from_pretrained(MODEL_NAME)
|
|
|
|
|
block = load_pretrained_block(MODEL_NAME, 0, config=config, torch_dtype=torch_dtype)
|
|
|
|
|
expected_dtype = resolve_block_dtype(config, torch_dtype)
|
|
|
|
|
assert all(param.dtype == expected_dtype for param in block.parameters())
|
