API Updates (#73)

* API Keys passed from Client instead of distro configuration

* delete distribution registry

* Rename the "package" word away

* Introduce a "Router" layer for providers

Some providers need to be factorized and considered as thin routing
layers on top of other providers. Consider two examples:

- The inference API should be a routing layer over inference providers,
  routed using the "model" key
- The memory banks API is another instance where various memory bank
  types will be provided by independent providers (e.g., a vector store
  is served by Chroma while a keyvalue memory can be served by Redis or
  PGVector)

This commit introduces a generalized routing layer for this purpose.

* update `apis_to_serve`

* llama_toolchain -> llama_stack

* Codemod from llama_toolchain -> llama_stack

- added providers/registry
- cleaned up api/ subdirectories and moved impls away
- restructured api/api.py
- from llama_stack.apis.<api> import foo should work now
- update imports to do llama_stack.apis.<api>
- update many other imports
- added __init__, fixed some registry imports
- updated registry imports
- create_agentic_system -> create_agent
- AgenticSystem -> Agent

* Moved some stuff out of common/; re-generated OpenAPI spec

* llama-toolchain -> llama-stack (hyphens)

* add control plane API

* add redis adapter + sqlite provider

* move core -> distribution

* Some more toolchain -> stack changes

* small naming shenanigans

* Removing custom tool and agent utilities and moving them client side

* Move control plane to distribution server for now

* Remove control plane from API list

* no codeshield dependency randomly plzzzzz

* Add "fire" as a dependency

* add back event loggers

* stack configure fixes

* use brave instead of bing in the example client

* add init file so it gets packaged

* add init files so it gets packaged

* Update MANIFEST

* bug fix

---------

Co-authored-by: Hardik Shah <hjshah@fb.com>
Co-authored-by: Xi Yan <xiyan@meta.com>
Co-authored-by: Ashwin Bharambe <ashwin@meta.com>

llama_toolchain/inference/quantization/loader.py

@@ -1,105 +0,0 @@
-# Copyright (c) Meta Platforms, Inc. and affiliates.
-# All rights reserved.
-#
-# This source code is licensed under the terms described in the LICENSE file in
-# the root directory of this source tree.
-
-# Copyright (c) Meta Platforms, Inc. and affiliates.
-# This software may be used and distributed in accordance with the terms of the Llama 3 Community License Agreement.
-
-import os
-from typing import Optional
-
-import torch
-
-from fairscale.nn.model_parallel.mappings import reduce_from_model_parallel_region
-from llama_models.llama3.api.model import Transformer, TransformerBlock
-from llama_toolchain.inference.api import QuantizationType
-
-from llama_toolchain.inference.api.config import (
-    CheckpointQuantizationFormat,
-    MetaReferenceImplConfig,
-)
-
-from termcolor import cprint
-from torch import Tensor
-
-
-def is_fbgemm_available() -> bool:
-    try:
-        import fbgemm_gpu.experimental.gen_ai  # noqa: F401
-
-        return True
-    except ImportError:
-        return False
-
-
-def swiglu_wrapper(
-    self,
-    x: Tensor,
-):
-    from .fp8_impls import ffn_swiglu
-
-    out = ffn_swiglu(x, self.w1.weight, self.w3.weight, self.w2.weight)
-    return reduce_from_model_parallel_region(out)
-
-
-def convert_to_quantized_model(
-    model: Transformer,
-    config: MetaReferenceImplConfig,
-    fp8_activation_scale_ub: Optional[float] = 1200.0,
-) -> Transformer:
-    if config.quantization.type == QuantizationType.bf16.value:
-        return model
-
-    elif config.quantization.type != QuantizationType.fp8.value:
-        raise ValueError("Only FP8 quantization is supported")
-
-    from .fp8_impls import Fp8ScaledWeights, load_fp8, quantize_fp8
-
-    checkpoint = config.checkpoint_config.checkpoint
-    # Move weights to GPU with quantization
-    if checkpoint.quantization_format == CheckpointQuantizationFormat.fp8_mixed.value:
-        cprint("Loading fp8 scales...", "yellow")
-        fp8_scales_path = os.path.join(
-            checkpoint.checkpoint_dir, f"fp8_scales_{get_model_parallel_rank()}.pt"
-        )
-        assert os.path.isfile(
-            fp8_scales_path
-        ), f"fp8_scales_path not found for rank {get_model_parallel_rank()}"
-        fp8_scales = torch.load(fp8_scales_path, weights_only=True)
-
-        for block in model.layers:
-            if isinstance(block, TransformerBlock):
-                if block.layer_id == 0 or block.layer_id == (model.n_layers - 1):
-                    continue
-
-                block.feed_forward.forward = swiglu_wrapper.__get__(block.feed_forward)
-                for key in ("w1", "w3", "w2"):
-                    param = getattr(block.feed_forward, key)
-                    param.weight = load_fp8(
-                        param.weight,
-                        fp8_scales[
-                            f"{block.layer_id}_feed_forward.{key}_{get_model_parallel_rank()}"
-                        ],
-                        fp8_activation_scale_ub,
-                    )
-    else:
-        cprint("Quantizing fp8 weights from bf16...", "yellow")
-        for block in model.layers:
-            if isinstance(block, TransformerBlock):
-                if block.layer_id == 0 or block.layer_id == (model.n_layers - 1):
-                    continue
-                block.feed_forward.forward = swiglu_wrapper.__get__(block.feed_forward)
-                for key in ("w1", "w3", "w2"):
-                    param = getattr(block.feed_forward, key)
-                    param.weight = quantize_fp8(
-                        param.weight,
-                        fp8_activation_scale_ub,
-                        output_device=torch.device("cuda"),
-                    )
-
-    for _, parameter in model.named_parameters():
-        if not isinstance(parameter, Fp8ScaledWeights):
-            parameter.data = parameter.to(device="cuda")
-    return model