From 0ba1bdd476c2079f1220904d5f2a217d78bdb64a Mon Sep 17 00:00:00 2001
From: 游雁 <zhifu.gzf@alibaba-inc.com>
Date: 星期五, 07 六月 2024 03:21:53 +0800
Subject: [PATCH] auto frontend

---
 funasr/models/llm_asr/model.py |   21 +++++++++++----------
 1 files changed, 11 insertions(+), 10 deletions(-)

diff --git a/funasr/models/llm_asr/model.py b/funasr/models/llm_asr/model.py
index 411b59d..e2880c8 100644
--- a/funasr/models/llm_asr/model.py
+++ b/funasr/models/llm_asr/model.py
@@ -468,7 +468,7 @@
         if len(speech_lengths.size()) > 1:
             speech_lengths = speech_lengths[:, 0]
 
-        batch_size = speech.shape[0]
+        batch_size, frames, _ = speech.shape
 
         # audio encoder
         encoder_out, encoder_out_lens = self.audio_encoder(speech.permute(0, 2, 1), speech_lengths)
@@ -476,14 +476,8 @@
         # audio_adaptor
         encoder_out, encoder_out_lens = self.audio_adaptor(encoder_out, encoder_out_lens)
 
-        input_ids[input_ids == -1] = 0
-        input_ids[input_ids == -100] = 0
-        if hasattr(self.llm.model, "embed_tokens"):
-            inputs_embeds = self.llm.model.embed_tokens(input_ids)
-        elif hasattr(self.llm.model.model, "embed_tokens"):
-            inputs_embeds = self.llm.model.model.embed_tokens(input_ids)
-        else:
-            inputs_embeds = self.llm.model.model.model.embed_tokens(input_ids)
+        input_ids[input_ids < 0] = 0
+        inputs_embeds = self.llm.model.get_input_embeddings()(input_ids)
 
         batch_size, token_num, dims = inputs_embeds.shape
         _, l, _ = encoder_out.shape
@@ -492,7 +486,7 @@
             inputs_embeds[batch_idx, fbank_beg_idx : fbank_beg_idx + l, :] = encoder_out[
                 batch_idx, :l, :
             ]
-
+        labels_ids[labels_ids == -1] = -100
         model_outputs = self.llm(
             inputs_embeds=inputs_embeds, attention_mask=attention_mask, labels=labels_ids
         )
@@ -505,6 +499,13 @@
             stats["acc"] = acc_att
 
         stats["loss"] = torch.clone(loss.detach())
+        stats["batch_size"] = batch_size
+        stats["batch_size_x_frames"] = frames * batch_size
+        stats["batch_size_real_frames"] = speech_lengths.sum().item()
+        stats["padding_frames"] = stats["batch_size_x_frames"] - stats["batch_size_real_frames"]
+        stats["batch_size_x_tokens"] = token_num * batch_size
+        stats["batch_size_real_tokens"] = attention_mask.sum().item()
+        stats["padding_tokens"] = stats["batch_size_x_tokens"] - stats["batch_size_real_tokens"]
 
         # force_gatherable: to-device and to-tensor if scalar for DataParallel
         if self.length_normalized_loss:

--
Gitblit v1.9.1