Fix issues during StableLM models conversion due to config.json changes

[aahouzi]

Type of Change

• Stability.ai did some parameter naming changes on config.json (cc: https://huggingface.co/stabilityai/stablelm-zephyr-3b/commit/8b471c751c0e78cb46cf9f47738dd0eb45392071) that caused some issues in the gguf conversion step (convert-hf-gguf.py):

[user1@dnp-test4 llama.cpp]$ python convert-hf-to-gguf.py $STABLELM
Loading model: a2eb1af48d006a3f6ec66a3a434979f4c817e2d0
Traceback (most recent call last):
  File "/home/user1/aahouzi/llama.cpp/convert-hf-to-gguf.py", line 1933, in <module>
    main()
  File "/home/user1/aahouzi/llama.cpp/convert-hf-to-gguf.py", line 1914, in main
    model_instance = model_class(dir_model, ftype_map[args.outtype], fname_out, args.bigendian)
  File "/home/user1/aahouzi/llama.cpp/convert-hf-to-gguf.py", line 50, in __init__
    self.model_arch = self._get_model_architecture()
  File "/home/user1/aahouzi/llama.cpp/convert-hf-to-gguf.py", line 285, in _get_model_architecture
    raise NotImplementedError(f'Architecture "{arch}" not supported!')
NotImplementedError: Architecture "StableLmForCausalLM" not supported!

Description

• This PR fixes the issue, by adapting to the new changes in Stability naming. After the change:

[user1@dnp-test4 llama.cpp]$ python convert-hf-to-gguf.py $STABLELM
Loading model: a2eb1af48d006a3f6ec66a3a434979f4c817e2d0
gguf: This GGUF file is for Little Endian only
Set model parameters
Set model tokenizer
gguf: Adding 100000 merge(s).
gguf: Setting special token type bos to 100257
gguf: Setting special token type eos to 100257
gguf: Setting special token type unk to 100257
Exporting model to '/home/user1/.cache/huggingface/hub/models--stabilityai--stablelm-2-1_6b/snapshots/a2eb1af48d006a3f6ec66a3a434979f4c817e2d0/ggml-model-f16.gguf'
gguf: loading model part 'model.safetensors'
output.weight, n_dims = 2, torch.bfloat16 --> float16
token_embd.weight, n_dims = 2, torch.bfloat16 --> float16
blk.0.attn_norm.bias, n_dims = 1, torch.bfloat16 --> float32
blk.0.attn_norm.weight, n_dims = 1, torch.bfloat16 --> float32
blk.0.ffn_down.weight, n_dims = 2, torch.bfloat16 --> float16
blk.0.ffn_gate.weight, n_dims = 2, torch.bfloat16 --> float16
blk.0.ffn_up.weight, n_dims = 2, torch.bfloat16 --> float16
...
blk.9.attn_k.bias, n_dims = 1, torch.bfloat16 --> float32
blk.9.attn_k.weight, n_dims = 2, torch.bfloat16 --> float16
blk.9.attn_output.weight, n_dims = 2, torch.bfloat16 --> float16
blk.9.attn_q.bias, n_dims = 1, torch.bfloat16 --> float32
blk.9.attn_q.weight, n_dims = 2, torch.bfloat16 --> float16
blk.9.attn_v.bias, n_dims = 1, torch.bfloat16 --> float32
blk.9.attn_v.weight, n_dims = 2, torch.bfloat16 --> float16
output_norm.bias, n_dims = 1, torch.bfloat16 --> float32
output_norm.weight, n_dims = 1, torch.bfloat16 --> float32
Model successfully exported to '/home/user1/.cache/huggingface/hub/models--stabilityai--stablelm-2-1_6b/snapshots/a2eb1af48d006a3f6ec66a3a434979f4c817e2d0/ggml-model-f16.gguf'

Expected Behavior & Potential Risk

• N/A

How has this PR been tested?

• N/A

Dependency Change?

• N/A

[aahouzi]

@cebtenzzre one test in CI fails with:

ERROR: Could not install packages due to an OSError: [Errno 28] No space left on device

[cebtenzzre]

one test in CI fails with:

Yeah, this is something @crasm added in #4585 but it's not important - the CI just doesn't have enough disk space to run the test.