From: Molly Sophia <redacted>
Date: Fri, 20 Dec 2024 09:44:58 +0000 (+0800)
Subject: convert : fix RWKV v6 model conversion (#10913)
X-Git-Tag: upstream/0.0.4488~120
X-Git-Url: https://git.djapps.eu/?a=commitdiff_plain;h=0a11f8b7b5c39fdf6e91ef9674bc68ff08681af7;p=pkg%2Fggml%2Fsources%2Fllama.cpp

convert : fix RWKV v6 model conversion (#10913)

* Enable --no-context-shift for llama-perplexity example

Signed-off-by: Molly Sophia <redacted>

* RWKV 6: Fix error in ggml_cuda_op_bin_bcast

Signed-off-by: Molly Sophia <redacted>

---------

Signed-off-by: Molly Sophia <redacted>
---

diff --git a/common/arg.cpp b/common/arg.cpp
index c3d66efb..deb11378 100644
--- a/common/arg.cpp
+++ b/common/arg.cpp
@@ -626,7 +626,7 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
         [](common_params & params) {
             params.ctx_shift = false;
         }
-    ).set_examples({LLAMA_EXAMPLE_MAIN, LLAMA_EXAMPLE_SERVER, LLAMA_EXAMPLE_IMATRIX}).set_env("LLAMA_ARG_NO_CONTEXT_SHIFT"));
+    ).set_examples({LLAMA_EXAMPLE_MAIN, LLAMA_EXAMPLE_SERVER, LLAMA_EXAMPLE_IMATRIX, LLAMA_EXAMPLE_PERPLEXITY}).set_env("LLAMA_ARG_NO_CONTEXT_SHIFT"));
     add_opt(common_arg(
         {"--chunks"}, "N",
         string_format("max number of chunks to process (default: %d, -1 = all)", params.n_chunks),
diff --git a/convert_hf_to_gguf.py b/convert_hf_to_gguf.py
index ecd69be6..76ab11eb 100755
--- a/convert_hf_to_gguf.py
+++ b/convert_hf_to_gguf.py
@@ -3065,6 +3065,9 @@ class Rwkv6Model(Model):
         if new_name.endswith("time_mix_w2.weight"):
             data_torch = data_torch.permute(0, 2, 1)
 
+        if new_name.endswith("time_mix_decay.weight") or "lerp" in new_name:
+            data_torch = data_torch.squeeze()
+
         rescale_every_n_layers = self.hparams["rescale_every"]
         if rescale_every_n_layers > 0:
             if new_name.endswith("time_mix_output.weight") or new_name.endswith("channel_mix_value.weight"):