py : fix StableLM conversion after config.json changes (#5703)

* Fix issues during StableLM models conversion

* Fix hard coded layer_norm_eps

* Support layer_norm_eps for LlavaStableLM

Co-authored-by: Jared Van Bortel <redacted>
* Add missing parenthesis

Co-authored-by: Jared Van Bortel <redacted>
* Support rotary_factor for LlavaStableLM

Co-authored-by: Jared Van Bortel <redacted>
* fix typo

* Add StableLMEpochForCausalLM for safety

Co-authored-by: compilade <redacted>
* Add StableLMEpochForCausalLM for safety 2

Co-authored-by: compilade <redacted>
---------

Co-authored-by: Jared Van Bortel <redacted>
Co-authored-by: Jared Van Bortel <redacted>
Co-authored-by: compilade <redacted>

diff --git a/convert-hf-to-gguf.py b/convert-hf-to-gguf.py
index 32d54b45f3325a9220c003af1b53e88267bc3ef2..ae30b2a76971a91bd97f5dc90182ec73bed2f976 100755 (executable)
--- a/convert-hf-to-gguf.py
+++ b/convert-hf-to-gguf.py
@@ -192,7 +192,7 @@ class Model:
             return RefactModel
         if model_architecture == "PersimmonForCausalLM":
             return PersimmonModel
-        if model_architecture in ("StableLMEpochForCausalLM", "LlavaStableLMEpochForCausalLM"):
+        if model_architecture in ("StableLmForCausalLM", "StableLMEpochForCausalLM", "LlavaStableLMEpochForCausalLM"):
             return StableLMModel
         if model_architecture == "QWenLMHeadModel":
             return QwenModel
@@ -253,7 +253,7 @@ class Model:
             return gguf.MODEL_ARCH.REFACT
         if arch == "PersimmonForCausalLM":
             return gguf.MODEL_ARCH.PERSIMMON
-        if arch in ("StableLMEpochForCausalLM", "LlavaStableLMEpochForCausalLM"):
+        if arch in ("StableLmForCausalLM", "StableLMEpochForCausalLM", "LlavaStableLMEpochForCausalLM"):
             return gguf.MODEL_ARCH.STABLELM
         if arch == "QWenLMHeadModel":
             return gguf.MODEL_ARCH.QWEN
@@ -1074,10 +1074,11 @@ class StableLMModel(Model):
         self.gguf_writer.add_embedding_length(hparams["hidden_size"])
         self.gguf_writer.add_block_count(block_count)
         self.gguf_writer.add_feed_forward_length(hparams["intermediate_size"])
-        self.gguf_writer.add_rope_dimension_count(int(hparams["rope_pct"] * (hparams["hidden_size"] // hparams["num_attention_heads"])))
+        rotary_factor = self.find_hparam(["partial_rotary_factor", "rope_pct"])
+        self.gguf_writer.add_rope_dimension_count(int(rotary_factor * (hparams["hidden_size"] // hparams["num_attention_heads"])))
         self.gguf_writer.add_head_count(hparams["num_attention_heads"])
         self.gguf_writer.add_parallel_residual(hparams["use_parallel_residual"] if "use_parallel_residual" in hparams else True)
-        self.gguf_writer.add_layer_norm_eps(1e-5)
+        self.gguf_writer.add_layer_norm_eps(self.find_hparam(["layer_norm_eps", "norm_eps"]))
 
 
 class MixtralModel(Model):