vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

tts : add OuteTTS support (#10784)

Browse source 
* server : add "tokens" output

ggml-ci

* server : output embeddings for all tokens when pooling = none

ggml-ci

* server : be explicit about the pooling type in the tests

ggml-ci

* server : do not normalize embeddings when there is no pooling

ggml-ci

* llama : add OuteTTS support (wip)

* wip

* extract features

* first conv

* group norm

* resnet conv

* resnet

* attn

* pos net

* layer norm

* convnext

* head

* hann window

* fix n_embd + remove llama.cpp hacks

* compute hann window

* fft

* spectrum processing

* clean-up

* tts : receive input text and generate codes

* clip : fix new conv name

* tts : minor fix

* tts : add header + minor fixes

ggml-ci

* tts : add matchematical constant

ggml-ci

* tts : fix sampling + cut initial noise

* tts : fixes

* tts : update default samplers

ggml-ci

* tts : text pre-processing

* tts : outetts-voc -> wavtokenizer-dec

* tts : remove hardcoded constants

ggml-ci

* tts : fix tensor shapes

* llama : refactor wavtokenizer tensors

ggml-ci

* cont

ggml-ci

* cont [no ci]

* llama : update WavTokenizer to non-causal attn

* llama : handle no-vocab detokenization

* tts : add Python example for OuteTTS (wip)

* tts : extend python example to generate spectrogram

ggml-ci

* server : fix rebase artifacts

* tts : enable "return_tokens" in Python example

ggml-ci

* tts : minor fixes

* common : support HF download for vocoder
This commit is contained in:
Georgi Gerganov 2024-12-18 19:27:21 +02:00 • committed by GitHub
parent 7bbb5acf12
commit 0bf2d10c55
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
19 changed files with 2509 additions and 532 deletions
Download patch file Download diff file Expand all files Collapse all files

70
gguf-py/gguf/tensor_mapping.py
View file

@ -42,6 +42,7 @@ class TensorNameMap:
"emb_ln", # nomic-bert
"transformer.norm", # openelm
"rwkv.blocks.0.pre_ln", # rwkv
"backbone.norm", # wavtokenizer
),
# Position embeddings
@ -60,6 +61,7 @@ class TensorNameMap:
"lm_head.linear", # phi2
"output_layer", # chatglm
"head", # rwkv
"head.out", # wavtokenizer
),
# Output norm
@ -80,6 +82,7 @@ class TensorNameMap:
"transformer.norm", # openelm
"model.norm", # nemotron
"rwkv.ln_out", # rwkv
"backbone.final_layer_norm", # wavtokenizer
),
# Rope frequencies
@ -90,6 +93,10 @@ class TensorNameMap:
MODEL_TENSOR.ROPE_FACTORS_LONG: (),
MODEL_TENSOR.ROPE_FACTORS_SHORT: (),
MODEL_TENSOR.CONV1D: (
"backbone.embed", # roberta
),
}
block_mappings_cfg: dict[MODEL_TENSOR, tuple[str, ...]] = {
@ -681,6 +688,8 @@ class TensorNameMap:
"encoder.block.{bid}.layer.1.DenseReluDense.wo", # t5
),
############################################################################
# TODO: these do not belong to block_mappings_cfg - move them to mappings_cfg
MODEL_TENSOR.ENC_OUTPUT_NORM: (
"encoder.final_layer_norm", # t5
),
@ -693,6 +702,67 @@ class TensorNameMap:
MODEL_TENSOR.CLS_OUT: (
"classifier.out_proj", # roberta
),
#############################################################################
MODEL_TENSOR.CONVNEXT_DW: (
"backbone.convnext.{bid}.dwconv", # wavtokenizer
),
MODEL_TENSOR.CONVNEXT_NORM: (
"backbone.convnext.{bid}.norm", # wavtokenizer
),
MODEL_TENSOR.CONVNEXT_PW1: (
"backbone.convnext.{bid}.pwconv1", # wavtokenizer
),
MODEL_TENSOR.CONVNEXT_PW2: (
"backbone.convnext.{bid}.pwconv2", # wavtokenizer
),
MODEL_TENSOR.CONVNEXT_GAMMA: (
"backbone.convnext.{bid}.gamma", # wavtokenizer
),
MODEL_TENSOR.POSNET_CONV1: (
"backbone.posnet.{bid}.conv1", # wavtokenizer
),
MODEL_TENSOR.POSNET_CONV2: (
"backbone.posnet.{bid}.conv2", # wavtokenizer
),
MODEL_TENSOR.POSNET_NORM: (
"backbone.posnet.{bid}.norm", # wavtokenizer
),
MODEL_TENSOR.POSNET_NORM1: (
"backbone.posnet.{bid}.norm1", # wavtokenizer
),
MODEL_TENSOR.POSNET_NORM2: (
"backbone.posnet.{bid}.norm2", # wavtokenizer
),
MODEL_TENSOR.POSNET_ATTN_NORM: (
"backbone.posnet.{bid}.norm", # wavtokenizer
),
MODEL_TENSOR.POSNET_ATTN_Q: (
"backbone.posnet.{bid}.q", # wavtokenizer
),
MODEL_TENSOR.POSNET_ATTN_K: (
"backbone.posnet.{bid}.k", # wavtokenizer
),
MODEL_TENSOR.POSNET_ATTN_V: (
"backbone.posnet.{bid}.v", # wavtokenizer
),
MODEL_TENSOR.POSNET_ATTN_OUT: (
"backbone.posnet.{bid}.proj_out", # wavtokenizer
),
}
# architecture-specific block mappings